Implementing Vmware Server Virtualization On Juniper Networks Infrastructure

Transcript

Implementation Guide Implementing VMware Server Virtualization on Juniper Networks Infrastructure Juniper Networks, Inc. 1194 North Mathilda Avenue Sunnyvale, California 94089 USA 408.745.2000 1.888 JUNIPER www.juniper.net Part Number: 801009-002 Aug 2008 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Table of Contents Introduction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Scope. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Target Audience. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Terminology and Concepts. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Server Virtualization. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Network Virtualization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 Design Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Juniper Networks Data Center Network Reference Architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Switching Design—Top-of-Rack . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Top of Rack—EX 4200 Series Switches . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 EX 4200 Virtual Chassis and Its Application for Virtualized Servers. . . . . . . . . . . . . . . . . . . . . . .11 High Availability at the Core and Access Layers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 VMotion Traffic Flow Considerations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 VMotion Traffic Across the Access and Core Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 VMotion Traffic in the Access Layer Within the Same Rack (128 Gbps Virtual Backplane) . . 15 VMotion Traffic in the Access Layer Across Multiple Racks (10 Gbps Uplink) . . . . . . . . . . . . 15 How VMotion Works and Best Practices. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Configuring VMware Connectivity to the Access Layer Interconnecting to the Core Layer . . . . . 18 Data Center Access Layer and Core Layer Connection Recommendations. . . . . . . . . . . . . . 19 VMware Server Connection Recommendations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Configuring the EX 4200 Access Switches. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Configuring the MX960 Core Ethernet Services Router . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Configuring the Virtual Switch on VMware ESX. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 General Recommended Steps for Configuring a Virtual Switch on VMware ESX . . . . . . . . . 22 Configuring VMware Connectivity at the Access Layer Within the Same Rack (128 Gbps Virtual Backplane). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 General Recommended Steps for Configuring the EX 4200 Access Port for VMotion. . . . . . 23 Configuring VMware Connectivity at the Access Layer Across Multiple Racks (10 Gbps Uplink). . 24 General Recommended Steps for Configuring Uplink Virtual Chassis . . . . . . . . . . . . . . . . . 24 Summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Appendix A Code Samples. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 EX 4200 Spanning Tree Configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 MX960 Configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Appendix B VMware Key Concepts and Terms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 About Juniper Networks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 2 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure List of Figures Figure 1. EX 4200 Series and MX960 Positioned in Access, Aggregation and Core Layers . . . . . . . . . . 4 Figure 2. VMware ESX 3.5 Server, VMware Virtualization Layers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Figure 3. VLAN Tagging—VGT, EST and VST Modes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Figure 4. Juniper Networks Data Center Reference Architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Figure 5. Connectivity Between Rack Servers and EX 4200 Series Switches. . . . . . . . . . . . . . . . . . . . . 9 Figure 6. EX 4200 Series Switch in the Application and Data Services Network. . . . . . . . . . . . . . . . . 10 Figure 7. Scaling Through EX 4200 Virtual Chassis Switch Top-of-Rack Configuration . . . . . . . . . . . . 11 Figure 8. EX 4200 Implemented as Top-of-Rack Switches with VMware Connection. . . . . . . . . . . . . 11 Figure 9. Example of High Availability and Load Balancing in Core and Access Layers. . . . . . . . . . . . 13 Figure 10. VMotion Traffic Flows Across the Access Layer (Virtual Chassis) and Core Layer . . . . . . . . 14 Figure 11. VMotion Traffic Flows in the Access Layer Within the Same Rack (128 Gbps Virtual Backplane). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Figure 12. VMotion Traffic Flow Across 10 Gbps Uplink to Different Racks. . . . . . . . . . . . . . . . . . . . . 15 Figure 13. Virtual Center/Production Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Figure 14. Interconnecting VMware ESX Server, EX 4200 Series Switch and MX960 Ethernet Services Router. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Figure 15. Connecting VMware ESX Server to Access and Core Layer . . . . . . . . . . . . . . . . . . . . . . . . 19 Figure 16. Example of VMware ESX Server NIC Teaming to the Access Layer. . . . . . . . . . . . . . . . . . 20 Figure 17. VMware ESX Virtual Switch Configuration. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Figure 18. VMware Connectivity Across 128 Gbps Virtual Chassis Backplane. . . . . . . . . . . . . . . . . . . 23 Figure 19. VMware Connectivity Across Multiple Racks (Using 10 Gbps Uplink). . . . . . . . . . . . . . . . . 24 Figure 20. Virtual Chassis Racks Across 10 GbE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Copyright ©2008, Juniper Networks, Inc. 3 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Introduction In today’s enterprise data centers, servers, applications and networking platforms are being consolidated. Network technology is being standardized to reduce costs, increase efficiencies and improve resource utilization. Virtualization is a key technology that meets these requirements. Applications are no longer bound to specific physical resources, but instead rely on a pool of CPUs, memory and network infrastructure services made available through virtualization. This paper provides various server and network virtualization design considerations typically associated with the data center environment. These design considerations are based on deploying Juniper Networks EX 4200 series switch with Virtual Chassis technology and the MX-series Ethernet Services Routers (ESR). We provide specific design considerations and implementation guidelines around VMware implementation in the data center network. We address specific High Availability (HA) scenarios with special attention given to the flexibility of the virtual chassis technology based architecture. Design and optimization steps address VMware’s ESX 3.5 virtual server network deployment, scalability and redundancy advantages. The implementation guidelines cover the following major topics: • Configuring Juniper Networks EX 4200 series switches and MX-series routers to integrate with VMware’s ESX virtual server for establishing a flat Layer 2 domain across the aggregated server pool • Configuring the EX 4200 Virtual Chassis technology to implement in the access layer • Configuring a 10 GbE uplink from an EX 4200 series switch Figure 1 shows an example of the EX 4200 series switches and MX960 ESRs topology in the access, aggregation and core layers. MX960-1 MX9 MX9 60 MX960-2 60 Core MX480 MX4 MX4 80 80 MX480 Aggregation Virtual Chassis EX 4 Serie200 s Access EX 4200-1 EX 4 Serie200 s EX 4200-2 EX 4 Serie200 s EX 4200-3 Figure 1. EX 4200 Series Switches and MX960 Positioned in Access, Aggregation and Core Layers 4 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Scope This paper presents design considerations, best practices and implementation guidelines for integrating VMware’s ESX 3.5 infrastructure into the enterprise network, in conjunction with deploying Juniper Networks EX 4200 series access switches and MX-series ESRs based on Juniper Networks data center network reference architecture. Target Audience • Security and IT engineers • Network architects Terminology and Concepts The following terms represent key items and concepts related to virtualization. These will serve as an aid in helping you understand virtualization as it applies to the design considerations and implementation guidelines presented in this paper. For further details concerning VMware-related concepts and additional documentation, see Appendix B, VMware Key Concepts and Terms. Server Virtualization Server virtualization is the masking of server resources, including the number and identity of individual physical servers, processors and operating systems, from server users. The server administrator uses a software application to divide one physical server into multiple isolated virtual environments. The virtual environments are sometimes called virtual private servers, but they are also known as partitions, guests, instances, containers or emulations. There are three popular approaches to server virtualization: the virtual machine model, the para-virtual machine model, and virtualization at the operating system (OS) layer, as shown in Figure 2. The virtual machine model (addressed in this paper) is based on the host/guest paradigm. This model permits dissimilar operating systems to run side-by-side on the same hardware by running under a virtual machine, which effectively tricks OS running under it into thinking it is running by itself on the entire native hardware. This approach allows the guest operating system to run without modifications. VMware and Microsoft Virtual Server both use the virtual machine model. To stay within the scope of this paper, we only address the concept of server virtualization in the virtual machine model. App App App App App Operating System Operating System Operating System Operating System Service Console VMware VIrtualization Layer x86 Architecture Disk CPU Memory NIC Disk Figure 2. VMware ESX 3.5 Server, VMware Virtualization Layers Copyright ©2008, Juniper Networks, Inc. 5 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Network Virtualization Network virtualization is the process of abstracting a single physical network, so as to permit its shared use by several logical entities—thus ensuring efficient utilization of that physical network. Network virtualization allows enterprises to segment the network using “virtual” objects, such as virtual LANs (VLANs), virtual routers, security zones, virtual route forwarding and virtual systems. These features can help enterprises take advantage of the efficiencies and security of network segmentation, without significantly increasing cost or network complexity and administrative burden. VLANs provide for logical groupings of stations or switch ports, allowing communications as if all stations or ports were on the same physical LAN segment. Confining broadcast traffic to a subset of switch ports or end users saves significant amounts of network bandwidth and processor time. VLANs are used in widely distributed computing environments as a means to identify and then segment traffic at a very granular level. Within Juniper Networks appliances, VLANs are used as a complement to virtual systems and security zones, segmenting traffic with a VLAN tag, and then routing that traffic to a security zone, a virtual system or a physical interface based on a security policy. Security Zones allow enterprises to divide the physical network into a series of virtual sections, so that they can establish various levels of trust and contain attacks. With Juniper Networks security zones, enterprises can define areas as specific as the finance department and wireless LAN users or as general as the untrust, trust and demilitarized zones (DMZ). Unlike some competitive security solutions that rigidly support untrust, trust and DMZs, each of the security zones in Juniper’s products can be customized to suit specific security needs. Multiprotocol Label Switching (MPLS) is a standards-approved technology for speeding up network traffic flow and making it easier to manage. MPLS involves setting up a specific path for a given sequence of packets, identified by a label put in each packet, thus saving the time needed for a router to look up the address of the next node in order to forward the packet. Label-switched path (LSP) is a sequence of hops (R0...Rn) in which a packet travels from R0 to Rn through label switching mechanisms. A label-switched path can be chosen dynamically, based on normal routing mechanisms or through configuration. In the enterprise core network, Juniper Networks recommends a two-layer aggregation network model as opposed to the conventional three-layer aggregation network model. By combining the EX 4200 series switch in the access layer and line-rate 10 GbE MX-series ESRs in the aggregation/core layers, we can collapse the switching tiers in the data center LAN. The EX 4200 series switch allows us to combine multiple racks and multiple rows into a single logical switch for administrative purposes. Uplinks from several racks now become optional, and this reduces the number of connection points required in the aggregation tier. The MXseries allows us to deliver line-rate 10 GbE at scale and full performance with all features turned on. This by itself requires fewer ports, and hence switches, compared to other switches in the marketplace. The overall reduction in the number of connection points in the aggregation tier allows the Juniper Networks architecture to reduce the total number of switches in the aggregation tier and, in several cases, reduce the number of tiers in data center LAN switching. From a network virtualization perspective, there are various technologies that provide data, control and management plane virtualization. A data plane virtualization example is a single physical interface to provide security to multiple network segments using 802.1q VLAN tagging. From a control plane virtualization perspective, multiple routing domains and protocol instances are other examples. A management plane virtualization example supports multiple logical firewall/VPN security systems that use virtual systems (VSYS) for true multi-department or multi-customer environments, such as large enterprises or service providers who offer managed security services all in a single device. Figure 3 illustrates VLAN tagging and emphasizes its three modes—Virtual Switch (internal to the hypervisor), EX 4200 series switch and Virtual Guest tagging. 6 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Operating System Operating System Operating System Operating System Operating System Operating System Operating System Operating System Operating System vSwitch vSwitch vSwitch Vlan tags applied in vSwitch Vlan tags applied in Guest EX-s eries v nic App v nic App v nic App v nic App v nic App v nic App v nic App v nic App v nic Port Groups assigned to a vlan App EX-s PortGroups set to vlan 4095 EX-s eries eries EX-Series switch applies vlan tags Figure 3. VLAN Tagging—VGT, EST and VST Modes NOTE: The dashed blue lines indicate the origin and direction of VLAN tagging. For additional VMware concepts and terms, see Appendix B VMware Key Concepts and Terms for definition of major terms used in VMware nomenclature discussed in this paper. Also visit http://www.vmware.com/products/vi/vc/vmotion.html. We will now show you how Juniper’s networking devices are leveraged by Juniper Networks reference architecture in the following design considerations section. Design Considerations Now let’s talk about how virtual networking and all of its major components—the virtual switch (internal to the hypervisor), NIC teaming, and VMotion apply to the Juniper Networks Data Center Reference architecture. For more information about VMware ESX 3 server virtualization, see Appendix B VMware Key Concepts and Terms. Juniper Networks Data Center Network Reference Architecture Figure 4 depicts a high-level view of Juniper Networks Data Center Network Reference Architecture. The core/ aggregation network tier connects to the Applications and Data Services tier which hosts all of the servers, databases and storage. The Edge Services tier provides connectivity to the Internet and WAN as well as edge security services, including VPN termination. Copyright ©2008, Juniper Networks, Inc. 7 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Priva Inter te W AN Edge Services WAN Edge WX/ WAN Acceleration WXC M series VPN Termination Gateway M series ISG ISG net Internet Edge Server Security Gateway ISG Network Services Intrusion Detection and Prevention Internet Access Gateway Core Network M serieXs EX 4 200 E SX e4 ie2s00 Serrie s IDP Core Aggregation Router NS Secure Access (SSL) Applications and Data Services S serieA s Core Firewall EX 4 200 E SX e 4e2s0 0 Serriie s IP Storage Network EX 4 200 E SX e 4e2s0 0 Serriie s EX 4 200 E SX e 42s0 0 Serriie es Internal Storage Network External Storage Network EX 4 200 E SX e 4e2s0 0 Serriie s Infrastructure storage Network Figure 4. Juniper Networks Data Center Reference Architecture 8 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure In this paper, we primarily discuss the connectivity options for the Applications and Data Services tier, as shown in Figure 4. In data center environments, servers are interconnected to access switches deployed within server racks. These access switches are often referred to as “top-of-rack” switches due to their location within the data center. Top-of-rack switching, implemented with an EX 4200 series switch, provides increased levels of availability because of multiple independent operating characteristics and physical power sources. Servers connect to two different physical switches, each part of a separate virtual chassis ring. Each ring in turn connects to the core network while using a loop detection and HA Layer 2 protocol. See Figure 5. Rack 1 Rack 2 EX 4 Serie200 s VC 1 EX 4 Serie200 s EX 4 Serie200 s VC 2 EX 4 Serie200 s Standby Active 40 Rack Mount Servers 40 Rack Mount Servers Figure 5. Connectivity Between Server Racks and EX 4200 Series Switches As shown in Figure 6, each internal and external applications network can be segmented into several sub networks. The servers that host these applications connect with 1 GbE links to the Juniper Networks EX 4200 series switch with Virtual Chassis technology. The EX 4200 series switch connects to the network core/aggregation layer through a 10 Gbps connection. Depending on the number of servers, multiple EX 4200 series switches might be required, as shown in Figure 6. Juniper Networks recommends dual homing the access layer switches (EX 4200 series) using Layer 3 with OSPF equal-cost multipath (ECMP) instead of using the Spanning Tree Protocol (STP) for deterministic behavior for minimal packet loss. Copyright ©2008, Juniper Networks, Inc. 9 Implementing VMware Server Virtualization on Juniper Networks Infrastructure M serieXs M serieXs Network Core 10GBASE-SR LAG Application and Data Services VC 1 VC 2 EX 4 Serie200 E X4s Serie200 s EX 4 Serie200 E X4s Serie200 s EX 4 Serie200 E X4s Serie200 s EX 4 Serie200 E X4s Serie200 s 10/100/1000BASE-T/TX Figure 6. EX 4200 Series Switch in the Application and Data Services Network NOTE: This recommendation applies except when the aggregated server pool over which you are running live migration does not lie entirely within a virtual chassis and you need to extend the Layer 2 domain to the MX960s. Switching Design—Top-of-Rack The “top-of-rack” EX 4200 series switch with Virtual Chassis technology configuration is the primary design consideration in a virtual chassis design for connecting virtualized servers. This section defines and discusses how this configuration applies to VMware virtualization and the Juniper Networks EX 4200 series switch. Top of Rack—EX 4200 Series Switches The Juniper Networks EX 4200 series switch eliminates the cable management headaches posed by “end-ofrow” chassis-based platforms. It also eliminates the need for multiple, commonly used “top-of-rack” stackable switches—an architecture traditionally used to provide a high availability design. In the following “top-of-rack” configuration, a pair of stackable EX 4200 series switches with Virtual Chassis technology typically sits atop a rack of servers (multiple servers can also be stacked in a “top-of-rack” formation). For redundancy purposes, each VMware ESX physical server in the rack supports two connections, one to each virtual chassis, as shown in Figure 7. In this configuration, EX 4200 series switches with Virtual Chassis technology are stacked on top of each other and are connected to the core/aggregation switches in a dual-homed manner, providing the necessary “always-on” link between users and critical backend VMware ESX physical servers. 10 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure 128 Gbps Virtual Backplane TOR Virtual Chassis #1 TOR Virtual Chassis #2 Vi Charstual sis Vi Charstual sis Vi Charstual sis Vi Charstual sis Vi Charstual sis Vi Charstual sis Vi Charstual sis Vi Charstual sis 10 GbE LAG to aggregation switches Figure 7. Scaling through EX 4200 Virtual Chassis Switch Top-of-Rack Configuration The Juniper Networks virtual chassis technology enables a data center to scale and add/stack as many EX 4200 series switches (up to 10 in a Virtual Chassis configuration) as needed to meet connectivity needs, while delivering true chassis-like functionality. For a more resilient design expanding port density, up to 10 EX 4200 series switches can be interconnected over a 128 Gbps backplane, thereby creating a single logical switch that supports up to 480 10/100/1000BASE-T ports and up to 40 GbE or 20 10 GbE uplink ports using the uplink module. For more information about the benefits of top-of-rack and end-of-row deployment, refer to the Juniper Networks Data Center LAN Connectivity Design Guide. EX 4200 Virtual Chassis Technology and Its Application for Virtualized Servers With the Juniper Networks EX 4200 series switches, IT gets the best of both worlds—deployment flexibility and the economics of stackable platforms combined with the advanced Layer 2 though Layer 4 functionality and the carrier-class reliability of a high-end chassis platform. While the redundant top-of-rack configuration, as shown in Figure 8, provides resiliency against equipment failures, traditional stackable switches without Virtual Chassis technology in general lack many of the HA characteristics associated with chassis-based switches, such as redundant power supplies, redundant cooling fans and fast, seamless failover. TOR VC#1 EX 42 00 EX 42 00 128 Gbps Virtual Backplane EX 42 00 GbE or 10 GbE uplink to another VC rack or Core/Aggregation layer TOR VC#2 EX 42 00 EX 42 00 128 Gbps Virtual Backplane EX 42 00 GbE or 10 GbE uplink to another VC rack or Core/Aggregation layer Rack 1 Rack 2 ... ... Rack 10 Figure 8. EX 4200 Series Switches Implemented as Top-of-Rack Switches with VMware Connection Copyright ©2008, Juniper Networks, Inc. 11 Implementing VMware Server Virtualization on Juniper Networks Infrastructure For core/aggregation switches that have thousands of Media Access Control (MAC) address associations, or with dense “end-of-row” access switch deployments greater than 1000 MAC addresses, software-based MAC address learning can have performance issues. The hardware-based MAC-learning MX-series products such as the MX960 are perfect for implementing at the core/aggregation layer in a two-tier data center model. The MAClearning rate of the MX960 is confirmed to be higher than 10,000 MAC-addresses per second. At the access layer, the number of attached MAC addresses is lower than the MAC-learning rate of EX switches (700 MPS for EX 4200). For example, if a 48-port Ethernet switch has 40 Gigabit Ethernet interfaces attaching hosts, each with eight virtual machines, the switch needs to learn only 320 MAC addresses. Usually, the MAClearning rate is important when one of the switches fails and there is no STP implemented in the network. However, with the EX 4200 series Virtual Chassis implementation, the content-addressable memory (CAM) tables are already populated on the failed over switch—hence relearning of MAC addresses is not required. For details concerning the Spanning Tree Protocol for the EX 4200 switch, see Appendix A, EX 4200 Series Spanning Tree Configuration and Spanning Tree Protocol in Layer 2/Layer 3 Environments. High Availability at the Core and Access Layers The core network is a key component in enabling HA in the data center network. By connecting all networks to the core network with full redundancy at the core, HA is achieved without added complexity and dependency on network protocols and convergence. Traditionally, adding HA required redesign of the network, whereas by using a core network approach and standards-based redundancy protocols such as VRRP (Virtual Router Redundancy Protocol), OSPF or other open-standard protocols for rapid such-second convergence, HA is provided at easier operational overhead. In addition to adding redundant devices, it is extremely important to ensure that the core data center devices support in-service operations such as hot-swap interfaces and software upgrades. In Juniper Networks data center network reference architecture, the core network is implemented with MX-series Ethernet services routers that provide a very high level of performance, HA and network virtualization features. At the access layer, Juniper Networks EX 4200 series switches offer redundant hot-swappable power supplies and a field replaceable fan tray. In addition, when two or more EX 4200 series switches are interconnected to form a Virtual Chassis configuration, the JUNOS™ software makes use of the multiple Routing Engines to deliver graceful Routing Engine switchover (GRES) as well as Layer 2 non-stop forwarding in the event of a Routing Engine failure. This ensures that network operations continue uninterrupted and that no critical routing data is lost following a primary Routing Engine failure. Primary and backup Routing Engines are automatically assigned by the JUNOS software, dictating an orderly transfer of control-plane functions. Figure 9 illustrates an example of the Juniper Networks EX 4200 access switch that provides HA and load balancing at the access layer (Virtual Chassis) along with the MX960 Ethernet services router which lies in the core layer. 12 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure MX9 MX9 60 60 Core Layer EX 42 00 Virtual Chassis EX 42 00 EX 42 00 Virtual Chassis EX 42 00 Access Layer (virtual chassis) Rack 1 Rack 2 Figure 9. Example of High Availability and Load Balancing in Core and Access Layers As illustrated in Figure 9, two redundant VMware ESX servers are placed in two separate racks and are interconnected to the data center infrastructure access layer, using two EX 4200 series switches with Virtual Chassis technology on each rack as “top of rack” switches, where each EX 4200 series switch connects with an uplink to the data center’s core/aggregation layer that runs the MX960 ESR. The key reason for this deployment is a redundant configuration (each virtual chassis, Rack 1 and Rack 2, backs up the other). VMware’s default mode of operation utilizes both links at any time, distributing the traffic across links based on network interface card (NIC) teaming and port group configuration. NIC teaming is a VMware feature where you can connect a single virtual switch (internal to hypervisor) to multiple physical Ethernet adapters. See Appendix B VMware Key Concepts and Terms for a detailed definition on NIC teaming. A team can share the load of traffic between physical and virtual networks among some or all of its members, and it can provide passive failover in the event of a hardware failure or a network outage. To configure for NIC teaming, see Configuring the EX 4200 Series Access Switches. Each VMware ESX server physical interface connects to a different EX 4200 Virtual Chassis configuration. See Appendix A Code Samples which provides the configuration code samples of the VMware networking relevant to this configuration and shows the load balancing characteristics that are associated with different configurations. Each EX 4200 Virtual Chassis configuration connects through an uplink to the MX960 core tier. The core tier MX960 links the Virtual Server Networks, the Virtual Center Network, and the VMotion Networks across all virtual chassis. The EX 4200 series switch configuration matching with the VMware port group setup is also discussed in the following configurations. VMotion Traffic Flow Considerations This section primarily covers the following three scenarios for design considerations that focus on establishing VMotion traffic flow: • VMotion traffic across the access and core layers • VMotion traffic in the access layer within the same rack (128 Gbps virtual backplane) • VMotion traffic in the access layer across multiple racks (10 Gbps uplink) Copyright ©2008, Juniper Networks, Inc. 13 Implementing VMware Server Virtualization on Juniper Networks Infrastructure On VMware enterprise servers such as the ESX 3.5 server, VMotion is a feature that supports the live migration of running virtual machines between physical servers. It is a solution to eliminate planned downtime. Most planned downtime is for hardware maintenance activities such as memory and storage upgrades, power supply replacements or fan replacements. If you have hardware monitoring agents in place, you sometimes get advance warnings of impending hardware failures, allowing you to preemptively bring down a server for repairs. VMotion lets your servers keep running right through periods of hardware maintenance by migrating virtual machines to other VMware ESX server hosts with zero downtime. Your virtual machine resource allocations are preserved when moved with VMotion, and VirtualCenter’s resource monitoring tools make it easy to identify hosts with adequate resources to receive migrated VMs and guarantee committed service levels. See the section, How VMotion Works and Best Practices for further details. In reality, VMotion can only happen on two identical physical servers, which means the VMware ESX server that applies the VMotion feature is for redundancy purposes only. The three scenarios illustrated in the Implementation Guidelines section of this paper introduce the concept of VMotion traffic flowing in an enterprise data center network. VMotion Traffic Across the Access and Core Layers As illustrated in Figure 10, we show redundant VMware ESX servers connecting to different TOR virtual chassis, #1 and #2. The VMotion traffic, represented by the orange arrow, flows across the EX 4200 series switches in the same VLAN through the access layer (EX 4200 access switches) and the core layer (MX960 ESRs). MX9 MX9 60 60 Core Layer Access Layer (virtual chassis) TOR VC#1 EX 42 00 Virtual Chassis EX 42 00 Virtual Chassis EX 42 00 VMotion Connection TOR VC#2 EX 42 00 VMotion Connection Rack 1 Rack 2 Figure 10. VMotion Traffic Flows Across the Access Layer (Virtual Chassis) and Core Layer 14 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure VMotion Traffic in the Access Layer Within the Same Rack (128 Gbps Virtual Backplane) In the following scenarios, we illustrate the design options and advantages using Juniper Networks EX 4200 series access switch (Virtual Chassis). In this case, the redundant VMware ESX server pairs interconnect to the same Virtual Chassis configuration through the 128 Gbps virtual backplane, as shown in Figure 11. In this design option, the VMotion traffic, represented by the orange arrow, flows only a short distance across the two different racks through the 128 Gbps Virtual Chassis backplane cable. Access Layer (virtual chassis) TOR VC#1 EX 42 00 Virtual Chassis EX 42 00 Virtual Chassis EX 42 00 VMotion Connection TOR VC#2 EX 42 00 VMotion Connection Rack 1 Rack 2 Figure 11. VMotion Traffic Flows in the Access Layer Within the Same Rack (128 Gbps Virtual Chassis Backplane) VMotion Traffic in the Access Layer Across Multiple Racks (10 GbE Uplink) In the following scenario (Figure 12), the redundant VMware server pairs in Rack 1 and Rack 2 connect to different virtual chassis systems, Rack 1 and Rack 2, connect to different Virtual Chassis racks through a 10 GbE uplink interface. In this design, the VMotion traffic, represented by the orange arrow, flows a distance of more than 3 meters across two different racks, within the same data center through the 10 GbE uplink interface with a fiber connection. Access Layer (Virtual Chassis) TOR VC#1 EX 42 00 10 GbE uplink VMotion Connection VMotion Connection TOR VC#2 EX 42 00 Rack 1 EX 42 00 10 GbE uplink EX 42 00 Rack 2 Figure 12. VMotion Traffic Flow Across GbE Uplinks to different Racks Copyright ©2008, Juniper Networks, Inc. 15 Implementing VMware Server Virtualization on Juniper Networks Infrastructure How VMotion Works and Best Practices Let’s explore how VMotion works in more detail. First, there are several configuration requirements that must be followed for VMotion to function properly: • VMotion is only supported by the ESX server hosts under VirtualCenter management and has to be communicating in the same Layer 2 domain. • A dedicated Gigabit Ethernet network segment is required between the ESX server hosts to accommodate rapid data transfers. • The ESX server hosts must share storage logical unit numbers (LUNs) on the same Storage Area Network (SAN) along with the virtual disk files for the virtual machines. For migration to occur, the ESX server host must also be contained in those shared LUNs (see Figure 13). • The processors on the ESX server hosts must be of the same type. For example, VMotion from a Xeon host to an Opteron host is not supported because the processor architectures are quite different. VM A esx01 VMFS esx02 SAN VirtualCenter Network Production Network Figure 13. Virtual Center/Production Network 1. We start with virtual machine A running on host ESX01. We want to move VM A to our second host, ESX02, so that we can perform maintenance on host ESX0. However, VM A has active user connections and network sessions that we wish to preserve. 2. The VMotion migration is initiated from the Virtual Center client or a Virtual Center scheduled task or VMware SDK script. The first step is to copy the VM A configuration file to host ESX02 to establish an instance of the VM on the new host. The virtual machine configuration file is simply a small text file listing the virtual machine’s properties. 3. Next, the memory image of VM A is copied to the target host running ESX02. The memory image can be quite large, so the dedicated Gigabit Ethernet network required by VMotion allows that copy to proceed at high speed. Immediately before the VM A memory image copy begins, VMotion redirects new memory write operations on host ESX01 to a separate location on ESX01. It also keeps track of these locations through a memory bitmap which will record all VM A memory updates that occur until the virtual machine is suspended on ESX01. In that way, the full memory image is read-only and static during the VMotion operation. 4. Because the virtual disk file for VM A is stored on a Virtual Machine File System (VMFS)-formatted SAN LUN mounted by both ESX01 and ESX02, we do not need to transfer that potentially large file. VMFS is VMware’s default storage system for virtual machine files on physical Small Computer System Interface (SCSI) disks and partitions. The multiple access feature of the VMFS file system enables this time-saving method. 16 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure 5. VMotion now suspends VM A on the ESX01 and copies the memory bitmap to ESX02. This step is the only one in which activity is interrupted, and that interruption is too short to cause connections to be dropped or users to notice. As soon as the memory bitmap is copied over, VMotion resumes VM A on its new home, ESX02. 6. VMotion sends a Reverse Address Resolution Protocol (RARP) packet to the production network EX 4200 series switch to inform it that the switch port to use for VM A has changed to the interface connecting to host ESX02. The EX 4200 series switch then updates its MAC table with the new MAC address and physical interface association. That preserves all network connections in the same VMotion Layer 2 broadcast domain to VM A. This step can be validated on the EX 4200 series switch through commandline interface (CLI) command: show ethernet-switching table. 7. Some modified memory pages might still reside on ESX01 after VM A has resumed—their locations being specified by the copied over memory bitmap. When VM A needs access to those pages, VMotion will “demand page” them over to ESX02. 8. VMotion completes the memory image transfer by background paging the remaining memory of VM A over to target host ESX02 and does a final commit of all modified memory pages to the full VM A memory image. Now VM A is back to using its full memory image in read/write mode. 9. The VMotion migration is now complete and we finish with a cleanup operation by deleting VM A from host ESX02. Implementation In this section, we illustrate and describe best practices for interconnecting VMware’s ESX 3.5 server to the data center infrastructure using two Juniper Networks EX 4200 series switches with Virtual Chassis technology on each rack as “top-of-rack” switches. In this implementation, each virtual chassis connects to the data center’s core/aggregation layer (tier) that consists of the MX960 ESRs (see Figure 14). Depending on your requirement for VMotion traffic flow, the implementation steps for the following three scenarios, as previously discussed in the Design Considerations section, are as follows: • Configuring VMware connectivity to the access layer interconnecting to the core layer • Configuring VMware connectivity at the access layer within the same rack (128 Gbps virtual backplane) • Configuring VMware connectivity at the access layer across multiple racks (10 GbE uplink) The implementation guidelines for each of the scenarios consist of the following three major configuration steps: 1. Configuring the VMware ESX server 2. Configuring the EX 4200 series access switch 3. Configuring the MX960 Ethernet services router Copyright ©2008, Juniper Networks, Inc. 17 Implementing VMware Server Virtualization on Juniper Networks Infrastructure MX9 MX9 60 60 Core Layer EX 42 00 Virtual Chassis EX 42 00 EX 42 00 Virtual Chassis EX 42 00 Access Layer (virtual chassis) Rack 1 Rack 2 Figure 14. Interconnecting VMware ESX Server, EX 4200 Series Switch and MX960 Ethernet Services Router Configuring VMware Connectivity to the Access Layer Interconnecting to the Core Layer When connecting VMware’s ESX servers to Juniper Networks EX 4200 series switches and MX960 ESRs, you need the following hardware components: • Four EX 4200 series switches with Virtual Chassis technology/configuration • Two MX960 ESRs • Two VMware ESX servers. • VMware ESX servers with VMotion licenses 18 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Data Center Access Layer and Core Layer Connection Recommendations As an example, Figure 15 illustrates the connection principles for connecting an EX 4200 series Virtual Chassis configuration to an MX960 ESR. The MX960 ESR connects to two Virtual Chassis clusters with the MX960 acting as the hub between the two Virtual Chassis configurations. If communication is required between the two Virtual Chassis configurations, then the traffic will traverse through the MX960 router. MX960-A MX9 MX960-B irb.71: 172.16.56.2 irb.72: 172.16.57.2 irb.73: 172.16.58.2 60 irb.71: 172.16.56.3 irb.72: 172.16.57.3 irb.73: 172.16.58.3 MX9 60 Vlan71, 72, 73 Iperf Client 172.16.59.5 ge-0/0/5 xe-10/1/0 xe-10/0/0 VIP: 172.16.56.1 172.16.57.1 172.16.58.1 ge-0/0/5 xe-10/0/0 xe-10/1/0 Vlan71: 172.16.56.0/24 Vlan72: 172.16.57.0/24 Vlan73: 172.16.58.0/24 Vlan71: 172.16.56.0/24 Vlan72: 172.16.57.0/24 Vlan73: 172.16.58.0/24 EX4200-C EX4200-D 192.168.6.90 EX4200-E 192.168.6.91 xe-0/1/0 EX 42 00 192.168.6.93 xe-0/1/0 EX 42 00 Virtual Chassis ge-0/0/12 EX4200-F 192.168.6.92 ge-0/0/1 ge-0/0/12 EX 42 00 ge-0/0/1 EX 42 00 Virtual Chassis ge-0/0/12 ge-0/0/12 Vlan71, 72, 73 Vlan73 Vlan71, 72, 73 Vlan73 Vlan71, 72, 73 E2 E2 E3 E4 Vlan71: 172.16.56.5 Vlan72: 172.16.57.5 VMServer-J OBM: 190.168.3.60 VMotion: 172.16.58.60 E3 E4 Vlan71: 172.16.56.6 Vlan72: 172.16.57.6 VMServer-K OBM: 192.168.3.61 VMotion: 172.16.58.61 Payload: VLAN 71, VLAN 72 VMotion: VLAN 73 Client: VLAN 75 EX switch mgmt: 192.168.6.0/24 VMware server mgmt: 192.168.3.0/24 Figure 15. Connecting VMware ESX Server to Access and Core Layer Copyright ©2008, Juniper Networks, Inc. 19 Implementing VMware Server Virtualization on Juniper Networks Infrastructure The following bulleted list represents the VLAN and IP addresses that Juniper Networks deployed in the solutions engineering lab, represented in Figure 15. • Virtual Machine Traffic: VLAN 71, VLAN 72 (Virtual Switch Tagging) • VMotion Traffic: VLAN 73 (EX Switch Tagging) • Iperf Client: VLAN 75 • VMware ESX server management network: 192.168.3.0/24 • EX 4200 series switch management network: 192.168.6.0/24 VMware Server Connection Recommendations • Bind two or more port groups to one or more NICs installed on each VMware ESX server. Connect across to different EX 4200 Virtual Chassis configurations, as illustrated in Figure 16. • Connect each virtual switch port group to one or more virtual machine guests. The following list the major configuration procedures: 1. Configuring the EX 4200 series access switches 2. Configuring the MX960 core ESR 3. Configuring VMware server network connectivity MX960-A EX4200-C EX4200-E Member ID: 0 Role: Primary Member ID: 0 Role: Primary Uplink module Uplink module VMNIC3 VMNIC4 Bond Port Group A 1 Virtual Switch 2 Port Group B 31 VLAN 71 32 VLAN 72 Virtual Machines ESX Server Host Figure 16. Example of VMware ESX Server NIC Teaming to the Access Layer 20 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Configuring the EX 4200 Series Access Switches To configure the EX 4200 series switch, we assume that VLAN tagging happens in virtual switches for the NIC teaming scenario. See Table 1 which explains in detail NIC teaming and VLAN tagging. To configure EX 4200 series switch trunk ports (for example, ge-0/0/12 to VMware server connections and xe0/1/0 to core layer) on one of the EX 4200 Virtual Chassis configurations for NIC teaming, set the VLANs and interfaces as follows: [edit] 1. set vlans vlan71 vlan-id 71 interface ge-0/0/12.0 2. set vlans vlan71 vlan-id 71 interface xe-0/1/0.0 3. set vlans vlan72 vlan-id 72 interface ge-0/0/12.0 4. set vlans vlan72 vlan-id 72 interface xe-0/1/0.0 5. set interfaces ge-0/0/12 description nic-teaming-port 6. set interfaces ge-0/0/12 unit 0 family ethernet-switching port-mode trunk 7. set interfaces ge-0/0/12 unit 0 family ethernet-switching vlan members [vlan71 vlan72] 8. set interfaces xe-0/1/0 description trunk-to-core 9. set interfaces xe-0/1/0 unit 0 family ethernet-switching port-mode trunk 10. .set interfaces xe-0/1/0 unit 0 family ethernet-switching vlan members [vlan71 vlan72] For more information about configuring EX-series interfaces, see Complete Software Guide for JUNOS for EXseries Software Configuring the MX960 Core Ethernet Services Router On the MX960 ESR, you can configure one or more bridge domains to perform Layer 2 bridging. A bridge domain is a set of logical ports that share the same flooding or broadcast characteristics. Like a VLAN, a bridge domain spans one or more ports across multiple devices. Thus, MX-series routers can function as Layer 2 switches, each with multiple bridging or broadcast domains that participate in the same Layer 2 network. You can also configure Layer 3 routing support for a bridge domain. Integrated Routing and Bridging (IRB) provides simultaneous support for Layer 2 bridging and Layer 3 IP routing on the same interface. IRB enables you to route packets to another routing interface or to another bridge domain that has a Layer 3 protocol configured. You configure a logical routing interface by including the irb statement at [edit interfaces] hierarchy level and include that interface in the bridge domain. For more information about how to configure a routing interface, see the JUNOS Network Interfaces Configuration Guide. NOTE: For an example of the software code configuration for the MX960, see Appendix A, MX960 Configuration. Copyright ©2008, Juniper Networks, Inc. 21 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Configuring the Virtual Switch on VMware ESX Refer to Figure 17 which shows the way EX 4200 series switches with Virtual Chassis technology are configured on the VMware ESX virtual server and review the general recommended steps that follow. Figure 17. VMware ESX Virtual Switch Configuration General Recommended Steps for Configuring a Virtual Switch on VMware ESX 1. Bind two or more port groups to one or more NICs installed on each VMware server. Connect across to different EX 4200 series switches with Virtual Chassis technology as illustrated in the Figure 16 and Figure 17. 2. Connect each virtual switch port group to one or more VM guest, as illustrated in Figure 17. Configuring VMware Connectivity at the Access Layer Within the Same Rack (128 Gbps Virtual Chassis Backplane) In this scenario, a network administrator can connect individual EX 4200 series switches together to form one unit and manage the unit as a single chassis, called a Virtual Chassis configuration. Up to ten EX 4200 series switches can be interconnected, providing a maximum of 480 Gigabit Ethernet ports. The port density increases as you expand the Virtual Chassis configuration. To connect the VMware ESX server to the access layer over the EX 4200 series 128 Gbps Virtual Chassis backplane, the following VLAN and NIC considerations were deployed in the Juniper Networks solutions engineering lab (see Figure 18). • VMs in VLANs 71 and 72 benefit from the VM NIC3 and VM NIC4 bond. The EX 4200 series Virtual Chassis configuration load balances egress traffic across the bonded virtual machine NICs through the hash of the source virtual NIC MAC address or a hash of the source and destination IP addresses. ((For IP load balancing, remember to use IP hash, configure a link aggregation group (LAG), and turn off Link Aggregation Control Protocol (LACP).) • The EX 4200 series switch with Virtual Chassis technology uses all virtual machine NICs in the bond. If a link failure occurs, the EX 4200 Virtual Chassis reassigns VM traffic to the remaining functional interfaces defined in the bond. 22 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure To take advantage of the higher Virtual Chassis bandwidth capacity and software redundancy features, interconnect at least two Juniper Networks EX 4200 series switches in a Virtual Chassis configuration. Begin with a default configuration, consisting of two EX 4200 series member switches interconnected with the two dedicated 64 Gbps Virtual Chassis ports, located at the rear panel, as shown in Figure 18. Dedicated Virtual Chassis Ports Rear view Front view SWA-0 Member ID: 0 Role: Primary Uplink module SWA-1 Member ID: 1 Role: Backup VMNIC3 VMNIC4 Bond Port Group A 1 Virtual Switch 2 Port Group B 31 VLAN 71 32 VLAN 72 Virtual Machines ESX Server Host Figure 18. VMware Connectivity Across 128 Gbps Virtual Chassis Backplane General Recommended Steps for Configuring the EX 4200 Access Port for VMotion Now that you have made the physical connectivity between VMware ESX and EX 4200 series access switches, you are ready to configure the VMotion interfaces (ge-0/0/1 in this case), set spanning tree edge (portfast) for the VMotion interface, and add the VMotion VLAN into the trunk port to the core layer (xe-0/1/0) by performing the following steps on both EX 4200 series switches with Virtual Chassis technology. [edit] 1. set vlans vlan73 vlan-id 73 interface ge-0/0/1.0 2. set vlans vlan73 vlan-id 73 interface xe-0/1/0.0 3. Set interfaces ge-0/0/1 description vmotion-port 4. Set interfaces ge-0/0/1 unit 0 family Ethernet-switching port-mode access 5. Set interfaces ge-0/0/1 unit 0 family Ethernet-switching vlan members vlan73 6. set interfaces xe-0/1/0 unit 0 family ethernet-switching vlan members vlan73 7. Configure the EX 4200 spanning tree edge for VMotion interface (portfast) • Set protocols RSTP interface ge-0/0/1.0 Begin with a default configuration, consisting of two EX 4200 member switches interconnected with the two dedicated 64-Gbps Virtual Chassis ports, located on the rear panel as illustrated in Figure 17. This figure depicts how to connect the two switches together using the Virtual Chassis ports. As the Virtual Chassis configuration expands, Virtual Chassis port connectivity changes. The topology of the Virtual Chassis ports varies based upon the requirements of the network administrator. For a detailed explanation, refer to part 6 in the document Complete Software Guide for JUNOS for EX-series Software. Copyright ©2008, Juniper Networks, Inc. 23 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Configuring VMware Connectivity at the Access Layer Across Multiple Racks (10 GbE Uplink) In this scenario, the EX 4200 Virtual Chassis technology can also be extended beyond a single rack. Switches located in multiple racks or in multiple data centers can be interconnected by installing the optional uplink module (EX-UM-2XFP) and by connecting the 10 GbE fiber uplink ports, as shown in Figure 19. Rack A Rack A Dedicated Virtual Chassis Ports Rear view Front view SWA-0 xe-0/1/0 SWA-1 xe-1/1/0 Rack B Member ID: 0 Role: Primary Uplink module Member ID: 1 Role: Linecard Uplink module Rack B Dedicated Virtual Chassis Ports SWA-2 xe-0/1/0 SWA-3 xe-3/1/0 VMNIC3 Member ID: 2 Role: Backup Member ID: 3 Role: Linecard VMNIC4 Bond Port Group A 1 Virtual Switch 2 Port Group B 31 VLAN 71 32 VLAN 72 Virtual Machines ESX Server Host Figure 19. VMware Connectivity Across Multiple Racks (Using 10 GbE Uplink) General Recommended Steps for Configuring Uplink for Virtual Chassis Technology To use the uplink ports for interconnecting member switches, you must explicitly configure the uplink ports as Virtual Chassis ports. This includes configuring the uplink ports of a standalone EX 4200 series switch as Virtual Chassis ports prior to interconnecting the new member switch with the existing Virtual Chassis configuration. See Figure 20. 24 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Rear view Rack A Dedicated Virtual Chassis Ports Front view EX 4200 SWA-0 xe-0/1/0 EX 4200 SWA-1 xe-1/1/0 Rack B Dedicated Virtual Chassis Ports Member ID: 0 Role: Primary Uplink module Member ID: 1 Role: Linecard Uplink module Rack B EX 4200 SWA-2 xe-0/1/0 EX 4200 SWA-3 xe-3/1/0 Member ID: 2 Role: Backup Member ID: 3 Role: Linecard g020011 Rack A Figure 20. Virtual Chassis Racks Across 10 GbE NOTE: For redundancy purposes, as shown in Figure 20, configure an uplink Virtual Chassis port in both switches SWA-0 and SWA-1.This example omits the specification of the member member-id option in configuring the uplink for SWA-0. The command applies by default to the switch where it is executed. To configure a Virtual Chassis arrangement across multiple switches, perform the following steps. 1. Configure the mastership priority of SWA-0 (member 0) to be the highest possible value (255). This ensures that it functions as the primary of the expanded Virtual Chassis configuration. Refer to Figure 20 when reviewing this example. [edit virtual-chassis] user@SWA-0#set member 0 mastership-priority 255 2. Prepare the members in rack B for interconnecting with the member switches in rack A by setting uplink Virtual Chassis ports for member 0 and member 1: user@SWA-2> request virtual-chassis vc-port set pic-slot 1 port 0 user@SWA-2> request virtual-chassis vc-port set pic-slot 1 port 0 member 1 3. Prepare the members in rack A for interconnecting with the member switches in rack B by setting uplink Virtual Chassis ports for member 0 and member 1: user@SWA-0> request virtual-chassis vc-port set pic-slot 1 port 0 user@SWA-0> request virtual-chassis vc-port set pic-slot 1 port 0 member 1 4. Configure Virtual Chassis in rack A: set virtual-chassis member 0 mastership-priority 255 (max). set virtual-chassis member 0 mastership-priority 128 (default). 5. Configure Virtual Chassis with 10 GbE uplink port: root@EX4200-C> request virtual-chassis vc-port set pic-slot 1 port 0 request virtual-chassis vc-port set pic-slot 1 port 0 member 1 Copyright ©2008, Juniper Networks, Inc. 25 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Summary Juniper Networks offers a unique value proposition for deploying server virtualization using VMware. The EXseries Virtual Chassis technology-based access layer can aggregate servers across multiple rows and multiple racks without deploying the Spanning Tree Protocol. This permits the use of superior Layer 3 connectivity (at no extra cost) from the access to the core/aggregation layer. We don’t know of any other vendor’s solution that can provide this same benefit. By following Juniper’s recommended implementation guidelines addressed in this paper, successful server virtualization can be accomplished in a cost-effective manner while still maintaining a highly scalable and reliable network architecture. Our suggested designs establish a complete solution for server virtualization by integrating VMware’s ESX 3.5 infrastructure into the enterprise network in conjunction with deploying Juniper Networks EX 4200 series switches with Virtual Chassis technology and MX-series 960 Ethernet services routers (both running on JUNOS software) based on the Juniper Networks reference architecture. Juniper Networks data center reference architecture simplifies the data center network infrastructure and provides a highly available and high-performance design that supports current as well as future applications while reducing risk and total cost of ownership. Appendix A: Code Samples The following code samples include configurations for enabling spanning tree on Juniper Networks EX 4200 series switches with Virtual Chassis technology and for placing interfaces into a bridge domain on the MX960 ESR. EX 4200 Spanning Tree Configuration The following configuration displays the configuration elements for enabling spanning tree on the EX 4200 series switch. [edit protocols rstp] interface ge-0/0/0.0 { cost 29999; } interface xe-0/1/0.0 { cost 29999; } MX960 Configuration The following configuration represents how to create Layer 2 switching interfaces on an MX960 ESR. [edit interfaces] + + flexible-vlan-tagging; + encapsulation flexible-ethernet-services; + unit 71 { + encapsulation vlan-bridge; + vlan-id 71; + } + unit 72 { + 26 xe-10/1/0 { encapsulation vlan-bridge; Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure + vlan-id 72; + + } } [edit interfaces ae0] + unit 71 { + encapsulation vlan-bridge; + vlan-id 71; + } + unit 72 { + encapsulation vlan-bridge; + vlan-id 72; + } [edit interfaces irb] + unit 71 { + family inet { + address 172.16.56.2/24 { + vrrp-group 1 { + virtual-address 172.16.56.1; + priority 190; + preempt; + accept-data; + } + } + } + } + unit 72 { + family inet { + address 172.16.57.2/24 { + vrrp-group 1 { + virtual-address 172.16.57.1; + priority 190; + preempt; + accept-data; + } + } + + } } [edit routing-instances ServicesSwitch bridge-domains] + VLAN71 { Copyright ©2008, Juniper Networks, Inc. 27 Implementing VMware Server Virtualization on Juniper Networks Infrastructure + domain-type bridge; + vlan-id 71; + interface xe-10/1/0.71; + interface ae0.71; + routing-interface irb.71; + } + VLAN72 { + domain-type bridge; + vlan-id 72; + interface xe-10/1/0.72; + interface ae0.72; + routing-interface irb.72; + } [edit routing-instances] + + instance-type virtual-router; + interface irb.71; + interface irb.72; + 28 VMotionTest { } Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Appendix B: VMware Key Concepts and Terms The following table alphabetically lists and defines the key virtual components that comprise VMware’s ESX 3.5 platform. For further details concerning VMware concepts, products and information, see VMware Virtual Networking Concepts and visit http://www.vmware.com/support/pubs/vi_pubs.html. Term/Concept Definition Load Balancing Load balancing allows you to spread network traffic from virtual machines on a virtual switch across two or more physical Ethernet adapters, giving higher throughput than a single physical adapter could provide. When you set NIC teaming policies, you have many options for creating load balancing. See VMware Virtual Networking Concepts for these load balancing options. NIC Teaming You can connect a single virtual switch to multiple physical Ethernet adapters using the VMware infrastructure feature called NIC teaming. A team can share the load of traffic between physical and virtual networks among some or all of its members, and it can provide passive failover in the event of a hardware failure or a network outage. You can set NIC teaming policies at the port group level. For further details concerning how to perform an EX-series trunk configuration for NIC teaming, see Configuring the EX 4200 Access Switches. NOTE: All physical switch ports in the same team must be in the same Layer 2 broadcast domain. Port Group A port group specifies port configuration options such as bandwidth limitations and VLAN tagging policies for each member port. Network services connect to virtual switches through port groups. Port groups define how a connection is made through the virtual switch to the network. In typical use, one or more port groups are associated with a single virtual switch. Port groups aggregate multiple ports under a common configuration and provide a stable anchor point for virtual machines connecting to labeled networks. Each port group is identified by a network label, which is unique to the current host. A VLAN ID, which restricts port group traffic to a logical Ethernet segment within the physical network, is optional. NOTE: You can create a maximum of 512 port groups on a single host. Virtual Center VMware infrastructure 3 provides a rich set of networking capabilities that integrate well with sophisticated enterprise networks. These networking capabilities are provided by VMware ESX server and managed by VMware VirtualCenter. VMware Virtual Center provides tools for building and maintaining your virtual network infrastructure. You can use VirtualCenter to add, delete and modify virtual switches and to configure port groups with VLANs and teaming. You can use the VirtualCenter roles feature to assign the permissions a network administrator needs to manage the virtual network. For further details, see http://www.vmware.com/vmtn/resources/826. Also see Managing VMware VirtualCenter Roles and Permissions. Virtual Ethernet Adapter Virtual Ethernet adapters are used by the virtual machine. Although there are many flavors of virtual Ethernet adapters, they primarily have the same basic attributes: • Contain their own MAC addresses and unicast/multicast broadcast filters • Are strictly Layer 2 devices Virtualization Copyright ©2008, Juniper Networks, Inc. Virtualization is a technique for hiding the physical characteristics of computing resources from the way in which other systems, applications or end users interact with those resources. This means making a single physical resource such as a server, an operating system, an application or a storage device appear to function as multiple logical resources; or making multiple physical resources such as storage devices or servers appear as a single logical resource. Virtualization also means making one physical resource appear, with somewhat different characteristics, as one logical resource. 29 Implementing VMware Server Virtualization on Juniper Networks Infrastructure Term/Concept Definition Virtual Machine A virtual machine (VM) is a software implementation of a machine (computer) that executes programs like a real machine. A virtual machine can be configured with one or more virtual Ethernet adapters, each of which each has its own IP address and MAC address. As a result, virtual machines have the same properties as physical machines from a networking standpoint. In addition, virtual networks enable functionality not possible with physical networks today. Virtual NIC The virtual network interface card (NIC) is a virtualized network interface that resides within the virtual machine and presents the same MAC interface that an actual physical NIC interface would provide. Multiple virtual NICs can be configured to bond with the same physical interface, allowing multiple VM guests to share that interface. Internally, a virtual NIC operates the same as a physical NIC. A virtual NIC when compared to a physical NIC has the following advantage: • Provides redundancy for enhanced availability Virtual Ports The ports on a virtual switch provide logical connection points among virtual devices and between virtual and physical devices. They can be thought of as virtual RJ-45 connectors. Each virtual switch can have up to 1,016 virtual ports, with a limit of 4,096 ports on all virtual switches on a host. VLAN Tagging VLANs provide for logical groupings of stations or switch ports, allowing communications as if all stations or ports were on the same physical LAN segment. Confining broadcast traffic to a subset of the switch ports or end users saves significant amounts of network bandwidth and processor time. • Virtual machine guest tagging (VGT mode)—You can install an 802.1Q VLAN trunking driver inside the virtual machine, and tags will be preserved between the virtual machine networking stack and external switch when frames are passed from or to virtual switches. The format for the header of a packet tagged in this way is shown in Figure 3. Use of this mode requires that the physical switch provide a trunk. • External switch tagging (EST mode)—You can use external switches for VLAN tagging. This is similar to a physical network, and VLAN configuration is normally transparent to each individual physical server. There is no need to provide a trunk in these environments. Virtual Switch Tagging (VST mode)—This is the most common configuration. In this mode, you provision one port group on a virtual switch for each VLAN and then attach the virtual machine’s virtual adapter to the port group instead of connecting the adapter to the virtual switch directly. The virtual switch port group tags all outbound frames and removes tags for all inbound frames. It also ensures that frames on one VLAN do not leak into a dif¬ferent VLAN. Using this mode requires that the physical switch provide a trunk. See Figure 3. For details on using VLANs with VMware infrastructure, see the white paper titled VMware ESX Server 3802.1Q VLAN Solutions, available from the VMTN Web site (http:// www.vmware.com/vmtn/). Also refer to the VMware Virtual Networking Concepts for additional information on VLANs in VMware infrastructure. VMware ESX 3.5 Server The VMware ESX 3.5 server is a host operating system dedicated to supporting virtual servers or virtual machines. The ESX host system kernel (VMkernel) controls access to the physical resources of the server shared by the virtual machines. The ESX host system ensures that the following four primary hardware resources are available to guest virtual machines: • Memory • Processors • Storage (local or remote) • Network Adapters 30 Copyright ©2008, Juniper Networks, Inc. Implementing VMware Server Virtualization on Juniper Networks Infrastructure Term/Concept Definition VMware Service Console VMware Service Console provides management component interfaces to the ESX server. The console is based on the Red Hat Enterprise Linux server, with unique privileges for and responsibilities to the ESX system. It also provides access to the ESX host through SSH, Telnet, HTTP and FTP, and provides authentication and system monitoring services. VMware Virtual Ports The ports on a virtual switch provide logical connection points among virtual devices and between virtual and physical devices. They can be thought of as virtual RJ-45 connectors. Each virtual switch can have up to 1,016 virtual ports, with a limit of 4,096 ports on all virtual switches on a host. The virtual ports in an ESX server provide a rich control channel for communication with the virtual Ethernet adapters attached to them. ESX server virtual ports operate in the following manner: • Know authoritatively what the configured receive filters are for virtual Ethernet adapters attached to them. This means no MAC learning is required to populate forwarding tables. •  Unlike physical switches, know authoritatively the “hard” configuration of the virtual Ethernet adapters attached to them. This capability makes it possible to set such policies as “guest can’t change MAC address,” because the virtual switch port can essentially know for sure what is “burned into ROM” (actually, stored in the configuration file, outside control of the guest operating system). For details on learning how virtual ports work, see VMware Networking Concepts. VMware Virtual Switch Virtual switches are the key networking components in VMware infrastructure 3.5. You can create up to 248 virtual switches on each ESX server 3 host. A virtual switch is “built to order” at run time from a collection of small functional units. These functional units support a modular design and make up the foundation that should be followed for future development. Additionally, this modular design is beneficial for VMware, and third party developers can easily incorporate modules to enhance the system in the future. They are: • Core Layer 2 forwarding engine • VLAN tagging, stripping and filtering units • Layer 2 security, checksum and segmentation offload units VMotion VMware’s VMotion technology, unique to VMware, leverages the complete virtualization of servers, storage and networking to move an entire running virtual machine instantaneously from one server to another. The entire state of a virtual machine is encapsulated by a set of files stored on shared storage, and VMware’s VMFS cluster file system allows both the source and the target VMware ESX server to access these virtual machine files concurrently. The active memory and precise execution state of a virtual machine can then be rapidly transmitted over a high speed network. Since the network is also virtualized by VMware ESX, the virtual machine retains its network identity and connections, ensuring a seamless migration process. VMware VMotion allows network administrators to: • Perform live migrations with zero downtime, undetectable to the user • Continuously and automatically optimize virtual machines within resource pools • Perform hardware maintenance without scheduling downtime and disrupting business operations • Proactively move virtual machines away from failing or underperforming servers For further details concerning VMotion and how it applies to the scenarios described in this paper, see the VMotion Traffic Flow Considerations section. For details on learning how VMotion works, see http://www.vmware.com/support/ pubs/vi/vc/vmotion.html. Also see VMware’s Knocking Out Downtime with Two Punches: VMotion & VMware HA. Copyright ©2008, Juniper Networks, Inc. 31 Implementing VMware Server Virtualization on Juniper Networks Infrastructure About Juniper Networks Juniper Networks, Inc. is the leader in high-performance networking. Juniper offers a high-performance network infrastructure that creates a responsive and trusted environment for accelerating the deployment of services and applications over a single network. This fuels high-performance businesses. Additional information can be found at www.juniper.net. CORPORATE AND SALES HEADQUARTERS Juniper Networks, Inc. 1194 North Mathilda Avenue Sunnyvale, CA 94089 USA Phone: 888.JUNIPER (888.586.4737) or 408.745.2000 Fax: 408.745.2100 www.juniper.net Copyright 2008 Juniper Networks, Inc. All rights reserved. Juniper Networks, the Juniper Networks logo, JUNOS, NetScreen, and ScreenOS are registered trademarks of Juniper Networks, Inc. in the United States and other countries. JUNOSe is a trademark of Juniper Networks, Inc. All other trademarks, service marks, registered trademarks, or registered service marks are the property of their respective owners. Juniper Networks assumes no responsibility for any inaccuracies in this document. Juniper Networks reserves the right to change, modify, transfer, or otherwise revise this publication without notice. 32 APAC HEADQUARTERS Juniper Networks (Hong Kong) 26/F, Cityplaza One 1111 King’s Road Taikoo Shing, Hong Kong Phone: 852.2332.3636 Fax: 852.2574.7803 EMEA HEADQUARTERS Juniper Networks Ireland Airside Business Park Swords, County Dublin, Ireland Phone: 35.31.8903.600 Fax: 35.31.8903.601 To purchase Juniper Networks solutions, please contact your Juniper Networks sales representative at 1-866-298-6428 or authorized reseller.