Using VMware High Availability And VMotion With .
Using VMware HighAvailability andVMotion withMicrosoft ExchangeServer 2007 VirtualMachines on Dell PowerEdge ServersTodd MuirheadKong YangDell Enterprise TechnologyCenterVirtualization SolutionEngineeringDell Enterprise Technology Centerdell.com/TechCenterOctober 2007
ContentsExecutive Summary . 3Introduction . 4Microsoft Exchange Server 2007 . 4VMware Infrastructure 3 . 4Dell PowerEdge Servers for Exchange 2007 . 5Test Configuration . 6Server Hardware . 6SAN Storage . 6Exchange Server 2007 Virtual Machines . 7Exchange Load Generator Testing Tool . 8VMotion Testing . 9VMware HA Testing. 12Conclusions. 14Acknowledgements . 15TablesTable 1: PowerEdge 2950 server configuration. . 6Table 2: Dell/EMC Storage . 7Table 3: Configuration of Exchange 2007 VMs used in testing. . 7Table 4: VMotion testing results. . 9Table 5: VMware HA testing results . 12FiguresFigure 1: Windows performance monitor graph of LoadGen task queue length for 1VM supporting1000 LoadGen heavy users during VMotion test. Each VMotion event can be seen as thebrief spikes in queue length. 10Figure 2: Windows performance monitor graph of LoadGen task queue length for 1VM supporting2000 LoadGen heavy users during VMotion test. Each VMotion event can be seen asspikes in queue length, with two much larger spikes reflecting the difficulty the large VM hadrecovering. . 11Figure 3: Performance monitor graph of task queue length and tasks completed per second for1VM supporting 1000 LoadGen heavy users during VMware HA testing . 13Talk Back --- Tell us how the Dell Enterprise Technology Center can help your organization better simplify,utilize, and scale enterprise solutions and platforms. Send your feedback and ideas toEnterprise Techcenter@dell.com or visit us at dell.com/TechCenter.October 2007Page 2Dell Enterprise Technology Center
Section1Executive SummaryWhen running applications in a virtualized environment, such as VMwareInfrastructure 3, there are new capabilities that are made possible by thevirtualization layer. In some cases, it is these new capabilities that make thecase for virtualization compelling enough for even performance intensive,mission critical applications like Microsoft Exchange Server. Dell engineerstested Microsoft Exchange Server 2007 with VMware Infrastructure’s VMotion(live migration) and VMware HA (high availability) and found that, when properlysized, the Exchange VMs continue to perform well during both a live migrationfrom one physical server to another and also during a recovery from a serverfailure.October 2007Page 3Dell Enterprise Technology Center
Section2IntroductionMicrosoft Exchange Server 2007 is a high performance application that is theemail server for many small and large organizations. As email use has grown tobecome pervasive and a critical communications tool for workers, InformationTechnology (IT) departments have also elevated the priority level for the emailserver. Keeping email available and performing at a high level is important to allIT departments.Another trend in the datacenter in recent years has been the increasing use ofserver virtualization to enable physical server consolidation and simplifiedmanagement. Initially server virtualization was viewed as a tool for test anddevelopment, but quickly was adopted as a means to consolidate underutilizedservers by running multiple Virtual Machines (VMs) on a single physical server.As the capabilities of servers have increased and server virtualization softwarehas matured some customers have begun to move more and more of theirservers onto VMs.In order to achieve additional cost and management benefits, it makes sense toconvert as many servers as possible into VMs. Microsoft Exchange Server 2007is not the first application that should be run in a VM, but it is an example of ahigh performance application that, when properly sized, can be configured in aVM. In order to provide the most benefit, an Exchange VM must be able to bemoved from one physical server to another without interruption to users. It mustalso be possible to recover from a physical server failure by bringing up anExchange VM on another physical server. As part of sizing and configuring anapplication into a VM it is important to consider the performance of thesevirtualization specific aspects.Performance characterization of this solution – Dell PowerEdge Servers, VMwareInfrastructure 3, and Exchange 2007 – was completed and detailed in a separatewhitepaper which is located at 20080147-Muirhead.pdf This paper extends that study using thesame test configuration and focusing on the VMotion and VMware HA aspects.Microsoft Exchange Server 2007Microsoft Exchange Server 2007 utilizes a 64-bit architecture and a new rolesbased setup to provide improved performance and capacity over previousversions of Exchange. Key new features include increased inbox access options,flexible deployment options, and additional high availability (HA) and disasterrecovery features.VMware Infrastructure 3VMware provides a server virtualization platform, referred to as VMwareInfrastructure 3, based on ESX Server and Virtual Center, that provides thefeatures of 64-bit VM support, VMotion (live migration) and VMware HA (highavailability).October 2007Page 4Dell Enterprise Technology Center
Dell PowerEdge Servers for Exchange 2007Dell PowerEdge servers provide a good platform for Exchange 2007 with 64-bitcapable processors, external high speed storage options, and a variety ofenterprise class support options. Dell engineers have done testing in order toprovide sizing and configuration guidance for running Exchange 2007 which isavailable as an online sizing tool and also in a whitepaper located athttp://www.dell.com/exchange.In addition to testing Exchange 2007, Dell engineers also test, validate andcertify VMware Infrastructure on Dell PowerEdge servers. Although Microsoftdoes not currently support Exchange 2007 in a VM1, some customers areconsidering this solution today to extend the benefits of virtualization to includetheir e-mail infrastructure as well.This whitepaper provides information about how Exchange 2007 VMs, runningon Dell PowerEdge servers, perform with VMware VMotion and VMware HAfeatures of VMware Infrastructure 3. Understanding how these features ofVMware Infrastructure affect the behavior of Exchange VMs is key in decidinghow to size and configure the VMs and their virtual environment.1. When considering the deployment of Microsoft Exchange Server 2007 on a virtualization platform such asVMware ESX Server, please refer to 0.aspxfor information about theMicrosoft support policy for Exchange 2007 in such a virtualized environment.October 2007Page 5Dell Enterprise Technology Center
Section3Test ConfigurationThe documentation of the test configuration has been organized into the areas ofserver hardware, SAN storage, Exchange Server VMs, and the test program LoadGen. This test was done with the same test configuration as that used forthe whitepaper that covered the performance characterization of Exchange 2007VMs.Server HardwareThe servers used in the testing were Dell PowerEdge 2950 two-socket rackbased systems, configured with quad-core processors and 16GB of RAM.Detailed system configurations are shown in Table 1. These systems wereinstalled with VMware ESX Server 3.0.1 and were the focus of the testing.Another 2950 was installed with Microsoft Windows Server 2003 and theExchange Load Generator 2007 tool from Microsoft to be used to simulatethousands of email users. Additionally, a PowerEdge 1855 server was used as aVMware Virtual Center management server/console.Dell PowerEdge 2950VirtualizationSoftwareVMware ESX Server 3.0.1CPU2 x 2.33 GHz Quad-Core Intel XeonE5345 with 8MB CacheMemory16 GB (8 x 2 GB FB DIMMs)Internal Disks2 x 146 GB 10K RPM SASNICs2x 10/100/1000 Mb/s (Internal)Disk ControllerPERC 5/iFiber Channel HBAQLogic QLE 2462 – dual port FC4PCI-ExpressRemoteManagementDell Remote Access Card, 5thGenerationTable 1: PowerEdge 2950 server configuration.SAN StorageStorage was provided by a Storage Area Network (SAN)-attached Dell/EMCCX3-80 fibre channel storage array. Each of the servers running ESX Serverwas attached to the SAN via a dual-port QLogic 2462 Host Bus Adapter. TheESX Server hosts are part of the same group in Virtual Center, and so need tohave access to the same storage in order to enable the VMotion and HAcapabilities. So a single Storage Group was created in Navisphere (themanagement tool for the CX storage) for both hosts and all the LUNs that wereused for this test. The storage components in the test are shown in Table 2.October 2007Page 6Dell Enterprise Technology Center
Controller1 Dell/EMC CX3-80Disk Enclosures3 Dell/EMC DAE3PDisks36 x 146 GB/ 15K RPMLUNs2 8-Disk RAID 1/0 LUNs for Mailbox Databases2 4-Disk RAID 1 LUNs for Logs2 5-Disk RAID 5 LUN for mailbox database backup1 HotSpare DiskTable 2: Dell/EMC StorageExchange Server 2007 Virtual MachinesThe virtual hardware configuration of the VMs was varied in the tests to adjust forthe number of users in the small, medium, and large scenarios. The same VMswere used for all the tests with the needed changes made via Virtual Center.The VM configurations were sized properly for the number of users each wassupporting. Table 3 summarizes how the VMs were scaled to support theincreasing user loads. The sizing was based on the whitepaper Dell Exchange2007 Advisor and Representative Deployments ( www.dell.com/downloads/global/solutions/Dell Exchange 2007 Advisor.pdf ) and the VMworld 2007presentation – Virtualizing Exchange 2007: The Final Frontier(www.delltechcenter.com/page/Exchange and VMware).Small VMMedium VMLarge VM5001000200048168 GB8 GB15 GBData Disks4816Log Disks248vCPU124vNIC111UsersStorage GroupsRAMTable 3: Configuration of Exchange 2007 VMs used in testing.All VMs were installed with Windows Server 2003 x64 Enterprise Edition andExchange Server 2007 Enterprise Edition. The Mailbox, Client Access, and HubTransport Exchange server roles were also installed on all VMs. In a standardproduction environment, it may be better to split out these different roles ontoseparate VMs. In order to keep the test environment simple and to keep thefocus of the test on how VMotion and HA function with Exchange VMs, all ofthese required roles were installed on all Exchange VMs. The mailbox role is thefunction that was primarily stressed in this test and has the highest requirementsfor I/O and RAM.The large VM was limited to 15GB of RAM although the physical server had16GB to allow the ESX kernel, service console, and other management to haveresources available. Initial testing was done with all 16GB assigned to the VM,but this caused the ESX server to not have enough resources during the VMotionand HA tests that were performed. This lack of available resources causedOctober 2007Page 7Dell Enterprise Technology Center
swapping to occur which greatly impacted the performance of the Exchange VM,although it did not become unavailable.Exchange Load Generator Testing ToolMicrosoft Load Generator 2007 (LoadGen) was used to stress the ExchangeVMs. LoadGen simulates a number of MAPI (Mail Application Program Interface)clients accessing their email on Exchange servers. Included with LoadGen areprofiles for medium and heavy workloads. The heavy workload user profile wasused for this testing which creates users with an average mailbox size of 225MB.In order to simulate users, LoadGen users perform 16 different types of tasksthat typical email users do during the course of a normal day. Examples arelogon, logoff, make appointment, send mail, and browse the calendar. LoadGenprovides results in terms of the average latency and 95th percentile latency ittakes these tasks to complete.In order to study the performance of Exchange while undergoing a VMotion orVMware HA recovery it was important to focus on the amount of interruption andtime to resume a normal level of performance. In order to measure this, theLoadGen task queue length was used as a barometer. Using the establishedaverage task queue length as a baseline, it was possible to observe any impacton Exchange availability as a rise in the task queue length. Measuring how longit took for this queue to return to a normal level also indicated how longExchange was affected by either the VMotion or VMware HA failover.October 2007Page 8Dell Enterprise Technology Center
Section4VMotion TestingIn order to test an Exchange VM during VMotion, a Java program was used tomove the VM running Exchange 2007 back and forth between two ESX serverhosts while under stress, actively serving a given number of LoadGen emailusers.Small, medium, and large Exchange 2007 VMs were tested, each supporting anumber of users and sized with the necessary resources. Please see Table 3 inthe previous section for VM configuration details.The 500, 1000, and 2000 user Exchange 2007 VMs were each tested withVMotion while servicing their respective number of LoadGen users. In order tomake the test more intensive than the typical environment, the VM was movedbetween servers with VMotion nine times in a two hour period. The amount oftime it took for each VMotion to complete was recorded. The Microsoft Windowsperformance monitor counters for the LoadGen task queue length were used tomeasure both the average and peak task queue length as an indicator of theimpact to availability that the VMotion events caused. Table 4 summarizes theresults.Small VMMedium VMLarge VMSmall VM 4x(Only 1 Moved)Users500100020002000Avg Time forVMotion(minutes:seconds)3:514:566:261:25Task Queue Avg2827414Task Queue PeakDuring VMotion2260129828Table 4: VMotion testing results.An additional test was done with four small VMs, each supporting 500 users for atotal of 2000 users. In this test only one of the four VMs was moved between theservers, but all of them were under stress. This test allowed for comparisonsbetween how a small VM performs during VMotion when it is the only VM andwhen it is one of many VMs. It also provided some data about how a 2000 userworkload can be moved with VMotion when it is in one big VM versus multiplesmaller VMs.The time for recovery after each VMotion is actually just a few seconds with thesmall and medium VMs. Figure 1 is a performance monitor graph that shows theLoadGen task queue length for the medium (1000-user) VM during the VMotiontest. Each of the nine VMotion events are visible as spikes at regular intervals.October 2007Page 9Dell Enterprise Technology Center
The graph shows how little impact each of the VMotion events has on the taskqueue length, with the queue returning to normal almost immediately.Figure 1: Windows performance monitor graph of LoadGen task queue length for1VM supporting 1000 LoadGen heavy users during VMotion test. Each VMotionevent can be seen as the brief spikes in queue length.The large VM exhibited different behavior than the small and medium VMs. Ittook the large VM much longer to fully recover following a VMotion event. Figure2 shows the performance monitor graph for the LoadGen task queue lengthduring the VMotion testing for the large VM.The higher and longer lasting spikes in the task queue length can be partiallyattributed to the doubling of users which also doubles the rate that tasks arebeing submitted. However Instead of the task queue peak doubling, it wasroughly ten times of what it should have been if it had just doubled. . Although 1GB of RAM was left unassigned to VMs and overall processor utilization wasalways below 50%, it appears that ESX Server may need additional resources.October 2007Page 10Dell Enterprise Technology Center
Figure 2: Windows performance monitor graph of LoadGen task queue length for1VM supporting 2000 LoadGen heavy users during VMotion test. Each VMotionevent can be seen as spikes in queue length, with two much larger spikesreflecting the difficulty the large VM had recovering.October 2007Page 11Dell Enterprise Technology Center
Section5VMware HA TestingThere are many options that can be used to provide high availability solutions forExchange Server 2007. Windows failover based clustering is probably one of themost well known methods. Exchange Server 2007 now also provides localcontinuous replication (LCR) and cluster continuous replication (CCR) that areapplication level solutions that can be used. In contrast to these solutions,VMware HA is a high availability option specific to virtualization. In order tounderstand how Exchange Server 2007 performs in a VMware Infrastructureenvironment, the performance of Exchange in terms of VMware HA must beincluded.Virtual Infrastructure 3 includes the VMware HA feature that allows for a VM orset of VMs running on a cluster of ESX servers to be restarted on another ESXserver if the server that it was running on fails. In order to test how ExchangeServer 2007 VMs would react to a sudden failure and subsequent restarting byVMware HA, a series of server failure simulation tests were run with small,medium, and large VMs that were under stress using LoadGen. Approximately10 minutes into the 2 hour LoadGen test, the server where the Exchange VMwas running was powered off. VMware HA would then restart the VM on theremaining node in the cluster. The results from this testing is shown in Table k Task Queue1674931458Time to Recover8 minutes15 minutes22 minutesTable 5: VMware HA testing resultsThe time to recover is measured from the time the ESX server was powered offuntil the Exchange VM is able to get the task queue length back to its normalaverage range for the given size of VM. With the VMotion testing the recoverytime was extremely short, but because VMware HA is dealing with a failure andnot a planned move, the time is significantly longer.Figure 3 is a Windows performance monitor graph for 1VM supporting 1000LoadGen heavy usersthat shows the task q
VMware Infrastructure 3 VMware provides a server virtualization platform, referred to as VMware Infrastructure 3, based on ESX Server and Virtual Center, that provides the features of 64-bit VM support, VMotion (live migration) and VMware HA (high availability).
2.7 VMware vCenter Support Assistant 22 2.8 VMware Continuent 23 2.9 VMware Hyper-Converged Infrastructure Kits 23 2.10 VMware Site Recovery Manager 23 2.11 VMware NSX 24 2.12 VMware NSX Advanced Load Balancer 28 2.13 VMware SD-WAN by VeloCloud 29 2.14 VMware Edge Network Intelligence 30 2.15 VMware NSX Firewall 30
the VMware Hybrid Cloud Native VMware management tools extend on-prem services across VMware Hybrid Cloud vRealize adapters allow "first class citizen" status for VMware Cloud on AWS Leverage same in-house VMware tools and processes across VMware Hybrid Cloud Support the cloud agility strategy of the organisation without disruption
Virtual VMware vCenter server enhancements: VMware vCenter Single Sign-On (SSO) and VMware vSphere Web Client, Server Appliance, App High Availability (HA), High Availability (HA), Distributed Resource Scheduler (DRS), and Big Data Extensions) VMware vSphere storage enhancements: Support for 62 terabytes (TB) of VMware Virtual Machine Disk
VMware also welcomes your suggestions for improving our other VMware API and SDK documentation. Send your feedback to: docfeedback@vmware.com. . , and can assist development of applications for VMware vSphere and vCloud. The user interface retains . In the VMware Developer Center, find the introduction page for VMware Workbench IS. At the .
8. Install VMware Fusion by launching the “Install VMware Fusion.pkg”. 9. Register VMware Fusion when prompted and configure preferences as necessary. 10. Quit VMware Fusion. Create a VMware Fusion Virtual Machine package with Composer 1. Launch VMware Fusion from /Applications. 2. Cre
VMware View 18 VMware Mirage 21 VMware Workspace 24 Summary 25 Chapter 2 VMware View Architecture 27 Introduction 27 Approaching the Design and Architecture 27 Phase I: Requirements Gathering and Assessment 28 Phase II: Analysis 29 Phase III: Calculate 30 Phase IV: Design 32 VMware View Server Architecture 33 VMware View Connection Server 34
VMware, Inc. 9 About ThisBook The Guest Operating System Installation Guide provides users of VMware ESX Server, VMware GSX Server, VMware Server, VMware ACE, VMware Workstation, and VMware Fusion information about installing guest operating systems in
astm a266/a266m-13 342, 345 astm a333/a333m-13 246, 249 astm a334/a334m-04a (2010) 246, 249 astm a485-14 628 astm a508/a508m-14 343, 346 astm a537/a537m-13 115, 125 astm a541/a541m-05 (2015) 342, 345 1a asme sa-508/sa-508m 342, 345 asme sa-541/sa-541m 342, 345 astm a508/a508m-14 343, 346 astm a541/a541m-05 (2015) 342, 345 1cr12 gb 1220-92 519, 522 gb 3280-92 438, 440 gb 4226-84 519 gb 4237-92 .
