Intermediary transaction http://www.aliyun.com/zixun/aggregation/6858.html ">seo diagnose Taobao guest cloud host technology Hall
Choosing a virtual machine that is just enough to have a number that is not too many is always a challenge for a given server. At first, it seemed easy to run a set of virtual machines and a physical server running only one operating system.
Burton Group analyst Chris Wolf said, however, it is really tricky to make sure that the hardware can support additional workloads because the type of software running in a virtual machine environment is almost limitless, and each application has a slightly different host operating system and hardware requirements.
Ian Scanlon, the IS operations manager at Computacenter, a data center and IT services company in London, which covers most of Europe, says that consolidating physical servers into virtual machines will save money, You cannot save too much money on hardware without reducing the application's performance and the risk of angering end users.
If you're going to put 5 virtual machines on one server, you're running 6 operating systems and all the applications, you have to step up your efforts to generally handle the situation and maintain the service levels, performance, and high levels of these applications, Scanlon said. We have to invest more money than we have identified during capacity planning.
According to Gordon Haff, a high-performance computing analyst at Illuminata, it is complex to predict in detail and accurately how well the server functions as a virtual machine host because of the different capabilities of the chipset to support virtual workloads and hypervisor.
Haff says the pressure on a virtual machine to cache the processor is greater than the pressure exerted by the physical server. The ability of the processor to transform between application requirements and virtual machine hypervisor requirements is different.
Andi Mann, an analyst at the Enterprise Management Association (EMA), said that both Intel and AMD specialize in the ability to support virtualization and Virtual server migrations in their circuits. A specified server may be configured with 2 and 8 processors, each with 2 to 8 processing cores. Without specific and hard analysis, it is almost impossible to predict how your particular server configuration can handle the different workloads of the software.
If you do not study the amount of work you intend to put on the server, even if you ask the manufacturer's technical representatives and sales representatives directly, it will not give you a specific answer. Massimo Re Ferre, senior it designer at IBM Systems and Technology division, says that while there are no rules to follow, some thumb laws can get you closer to finding weaknesses and knowing where and how to strengthen them. First, you can add three to five virtual machines for each kernel of the new Intel or AMD processor.
This is more optimized than the Scanlon outlook. Scanlon said he placed five or six virtual machines on a single server. If these applications are resource-intensive databases or ERP applications, he runs only two applications.
Mann says the bottom line is that the less the better. Putting too much emphasis on consolidation inevitably leads to poor performance and user dissatisfaction. The second thumb law proposed by Re Ferre is: Add 2GB to 4GB memory for each kernel of the new processor. This approach is consistent with the Scanlon assessment. He runs 48GB of RAM on every high-end blade server. ' Once we've added the memory, we don't have any performance problems, ' he said.
Other issues to be concerned about:
First of all, don't forget the plumbing. Owning many servers has many input/output requirements. Make sure you have enough connections to connect to the back-end storage and network to accommodate these input/output requirements, Haff said.
Second, build a fence. Virtual machines are easy to launch and difficult to see, so it is common for virtual machines to spread (there are too many unused virtual machines running). Scanlon says that undoing all unused servers and the hard disk space they occupy allows Computacenter to reclaim a lot of resources so that the company can postpone significant upgrades to the next budget cycle.
Third, Wolf says, use all the tools available to let you know your settings in detail.
VMware offers VMware ' s vcenter Capacityiq 1.0 and Microsoft provides its "assessment and calculates Toolkit for Hyper-V" tool to help guide users. However, both tools are not good at setting up environments that contain VMware and Hyper-V.
Many Third-party tools do contain the contents of these two versions. HP has a dedicated tool for VMware and a tool specifically designed for Hyper-V. Mann says companies that are not close to VMware and Microsoft can provide more independent assessments.
These tools include Novell's Powerrecon, Cirba, Inc. 's Datacenter Intelligence and Akorri Balancepoint, Vkernel capacity Modeler. To learn more about how a customer can use the Cirba tool, see the security study at the cio.com website, "How intelligently the underwriter's lab plans virtualization".
Re Ferre said, finally, it is impossible to build a detailed data graph of the virtual machine host you require and expect. However, given the time and money you need to do this yourself or the cost of paying a professional assessment, it may be cheaper to buy more powerful hardware than you actually need, and you can run a server with a capacity of 70% because it will not be overly nervous at 95% capacity.