Source: Intel preso at Vmworld.Workingclosely with Vmware and our OEMs, VMmark performs very efficiently on Xeon 5500 series.As you can see all the major OEMs has published Vmmark scores on Nehalem based servers – super efficient virtualization is one of the requirements for cloud computing; Xeon 5500 series is the platform of choice
This is another example of how VMware enables higher performance from physical hosts.
The good news with SQL server consolidation – and Oracle consolidaiton – is that you can realize very significant cost savings not just on the infrastructure side, but also on the licensing side. Today, most SQL / ORCL database licenses are used at just 5% - 10% on average. By consolidating, you not only increase infrastructure utilization, but also license utilization. In this example, we are going from 8 individual DB each on a dedicated host, to a couple of shared host running 8 DB instances. With Enterprise edition, just need to license physical processors, so you effectively consolidate your licenses in addition to servers.This leads to a very significant reduciton in licensing costs, especially since software licenses tend to dominate the cost structure for SQL deployments.
Failback requires some setup _after_ fail-over is performed.Unique featuresSite Recovery Manager2 main featuresAutomated DRProof to Audit or Business that IT has done the DR properly with minimal hit on IT resources.It is possible because we have a standard API that many storage vendors have written to. They allow us to control their array (e.g. discover LUN, stop replication)Global ViewBrowse and Search across multiple data center. Scale up to 10 data centers worldwide.
Technically, you maybe able to hit 120 VM or more per cluster. But from best practice point of view, we don’t recommend such high ratio. 120 VM in 7 hosts (1 is assumed to be unavailable) is > 17:1 ratio
Infra VMs (vCenter, Update Manager, etc) connect to Management LAN. Since they are on the same physical NIC, we need to create VLAN.
In most customer environments, early virtualization deployments often focus on virtuliazation “candidates” that are considered lower risk – perhaps file and print servers or applications that are not considered mission critical. As customers seek to expand the virtualization footprint, by definition they want to virtualize business-critical apps – and vCenter AppSpeedis an excellent tool to help them achieve that with confidence. There are two main use cases for AppSpeed: 1. Guarantee application SLA for virtualized multi-tier apps 2. Perform an “Assured Migration”Guarantee application SLAAppSpeed allows you to continually monitor the performance of an application ensuring SLAs are met. The product automatically discovers and maps applications and components all the way down to specific database tables…even if the database or any app tier is still on a physical system. The product will then monitor real-time user end user experience for all transactions – breaking down performance into its component parts (e.g. network, application and database latency). Any time an application does not meet its assigned threshold, the user is alerted, and can quickly analyze the root cause of the problem. Users can quickly identify and dive into the application’s dependencies to pin-point the source of the problem. Perform “Assured Migration”One of the top inhibitors to Virtualizing applications is the concern expressed by app owners that applications may not perform as well in virtualized environments. The answer is to measure performance on the physical app pre virtualization, and again after it has been virtualized. In both use cases, AppSpeed, provides a shared view of the application that be understood by both IT and application owners, engendering better communication about application performance overall, and reducing fingerpointing and enable faster troubleshooting and resolution if and when trouble arises. In short, With vCenter AppSpeed, customers can be sure that their business critical applications will run as well or better in virtual mode.
The Trend Is Clear…<br />% of customers running appsin productionon VMware<br />56%<br />53%<br />50%<br />41%<br />34%<br />36%<br />27%<br />24%<br />MS Exchange<br />MS SharePoint<br />MS SQL<br />Oracle<br />Middleware<br />Oracle<br />DB<br />IBM WebSphere<br />IBM DB2<br />SAP<br />Source: VMware customer survey, September 2008, sample size 1038<br />Data: Within subset of VMware customers running a specific app, % that have at least one instance of that app in production in a VM<br />In a recent Gartner poll, 73% of customers claimed to use x86 virtualization for mission critical applications in production<br />Source: Gartner IOM Conference (June 2008)<br />“Linux and Windows Server Virtualization Is Picking Up Steam” (ID Number: G00161702)<br />
VMware VMmark<br />Use VMmark as the basis for CPU selection<br />It is the official benchmark for VMware, and it uses multiple workload<br />Other benchmark are not run on vSphere, and typically test 1 workload<br />Use it as a guide only<br />You need head room and HA.<br />How it’s done<br />1 Tile = 6 VM. Each is 1 vCPU.<br />MS Exchange, MySQL, Apache, J2EE, File Server, Idle VM<br />Result page: <br />www.vmware.com/products/vmmark/results.html<br />VMmark 1.1 results is directly comparable to VMmark 1.0 results. The underlying virtual hardware definitions and load levels for each workload have not changed.<br />1 Tile = 6 VM<br />Each 2 CPUs<br /> 2GB RAM<br />
Sample Results (as at Feb 2010. Major update in May)<br />VMware does not recommend 1 OEM over the other. <br />I use 1 OEM here so we don’t compare between OEM<br />Look at this number<br />This number is when comparing with same #Tiles<br />Opteron 8439, 24 cores<br />Xeon 5570, 8 cores<br />Opteron 2435, 12 cores<br />Xeon 5470, 8 cores<br />This tells us that DL380 G6can run 17 Tiles, at 100% utilisation.<br />Each Tile has 6 VM, but 1 is idle.<br />17 x 5 VM = 85 active VM in 1 box!<br />At 70% ESX host utilisation, that’s around 59 VM.<br />DL 385 G6 runs 11 tiles.<br />The 2 boxes have similar date submission, around mid 2009.<br />
Xeon 5500 improvement over previous generation<br />Confidential<br />8<br />Around 2:1 improvement<br />Details<br />Servers were Dell M600 and Dell 2950<br />Storage were a few Dell EqualLogic PS5000XV iSCSI arrays. <br />MS Exchange Load Generator (LoadGen) was used to run the tests. <br />The VM had 10GB of RAM and installed with Windows 2008 x64 Enterprise Edition and the Exchange mailbox role. <br />
Designing for Tier 1<br />How does your Architecture…..<br />lend itself for troubleshooting?<br />ensure high availability?<br />integrate with DR?<br />integrate with security?<br />minimise risk during vSphere update/upgrade?<br />encourage the expert own their respective area? Storage, Network, Server<br />help Application team monitor/troubleshoot?<br />Is vSphere alone sufficient?<br />Products from VMware<br />Products from non VMware that deliver superior technology<br />Next few slides show an example of an architecture.<br />
vCenterAppSpeed<br /><ul><li>Auto-discover all infrastructure across all tiers, both virtual and physical
Monitor application performance against SLAs at the transaction level
Root cause analysis and faster troubleshooting</li></ul>Application Performance Monitoring<br />
Value add over vSphere<br />AppSpeed provides application-awareness to vSphere.<br />Answer questions like:<br />What applications are running in my data center?<br />Are the users getting good response times?<br />When the application is slow, is it at all my VMware fault? Where is the bottleneck? <br />Am I meeting my SLA?<br />What are the different transactions a user makes on an application and how are they performing?<br />What is the relationship between application performance and resource utilization?<br />
Assured Migration<br /> Migrate applications from physical to virtual (“Assured Migration”)<br />Measure post virtualization performance <br />Develop pre-virtualization end-user application baseline<br />Measuring Physical App<br /><ul><li>Install a probe and ESX and direct a spanport to vSwitch
The “people” side <br />Knowledge<br />Design knowledge<br />Troubleshooting knowledge<br />Virtual environment is easy to install & configure. Troubleshooting is a different ball game.<br />Practice helps.<br />Read the logs (fastest way to put your kids to sleep )<br />VCP & VTSP are Level 100. Consider Advance VCP.<br />Process<br />vCOE<br />Proper Support<br />TAM<br />Business Critical Support<br />
Design Knowledge<br />Design Process Overview<br />ESX/ESXi Host Design<br />vSphere Virtual DC Design<br />vSphere Network Design<br />vSphere Storage Design<br />Virtual Machine Design<br />Management and Monitoring Design<br />
Common Questions<br />What’s in it for me?<br />Is my application supported?<br />Who owns Vmware, Windows or UNIX?<br />Who’s going to pay for these LUNs?<br />Trunks into a server? No way!<br />How are we going to secure this environment?<br />How do we monitor and who do we call?<br />
Business Critical Support<br />Designated Support Team<br />Direct Routing to Senior Level Engineers<br />Customer Site Profile<br />Proactive Health Check<br />Root Cause Analysis<br />Support for Migration and Updates<br />Monthly Support Review Meeting<br />http://www.vmware.com/support/services/bcs.html<br />
Common Questions<br />How do we charge for this?<br />What should we track, and how do we do it?<br />You mean it automatically changes?<br />What kind of changes are there and what are the risks?<br />How do we manage releases?<br />When do we expand?<br />How do we react to alerts?<br />
Virtualization Capability Assessment<br />Before doing Tier 1….<br />take an honest look at your organisation capability for Tier 1? <br />no need to be long & detailed assessment.<br />
VCA Methodology - Assessing the VI Ecosystem<br />Does the environment have the appropriate number of staff and/or skill set to implement, operate and manage the VMware Infrastructure?<br />Do the appropriate procedures exist to allow the organization to effectively manage the VMware Infrastructure at the required service level?<br />Do the necessary tools exist to support the processes and operate the virtualization environment?<br />Can the current VMware Infrastructure design (represented in the existing implementation) provide the desired service levels or operational behavior (i.e. automation, etc.)?<br />
Criteria Within Each Service Phase<br />People<br />Process<br />Technology<br />Architecture<br />
Application<br />Review the Ref Arch<br />Desktop<br />MS Exchange 2007<br />SAP<br />Do Application-sizing, not physical sizing.<br />1000 SAP SD users, not 8 core Xeon 55890<br />Go Horizontal if possible.<br />Consolidation Ratio may be lower<br />