The next-generation enterprise data center presents four major development trends including virtualization
Bill Fife, technical manager of Wholesale Electric Supply Co in Houston, said that some people are very accustomed to an application, a server and a method. This allows you to lock in a way of thinking, it is difficult to fully understand the new generation of data centers. Fife said, however, now has thin copy replay and synchronization with disaster recovery sites, and virtual machines can move files between data storage and multiple data storage on a server, and can also add network adapters, You really have to sit down and think about how you want to run your business and remember that you have some options. You should not be limited to one road. You can follow this path today and change direction tomorrow. Here are four main trends in data centers: Trend 1: I / O virtualization At Wholesale Electric Supply, Fife is leveraging virtualized I / O capabilities, which is one of several important technologies that shape a new generation of data centers. I / O virtualization (also called I / O aggregation) can separate the interconnection of 10 Gigabit Ethernet InfiniBand or Ethernet links. Xsigo Systems' virtual I / O Director uses the former technology, and Cisco's Nexus 5000 and 7000 switches use the latter technology. Logan Harbaugh, a member of the Network World Lab Alliance and an independent analyst, said that in either case, as long as the connection is established, you can get the number of virtual Ethernet and Fibre Channel connections you want. These architectures are similar, and although there is a limit to how many modifications you can make, they still provide some level of functionality. I / O virtualization simplifies the hardware status of the data center, while providing flexibility while significantly reducing the number of connections between various devices. Consider the best-practice recommendations from VMware, each virtual machine is assigned a 1G port. With a relatively new 24-core server, in theory a server can run at least 24 virtual machines, perhaps as many as 50 virtual machines. This means that 50 1G ports are needed. Realistically, even if you can use 6 4-port Ethernet cards, you can only support 24 virtual machines. The best thing to do about I / O virtualization is that everything shares an InfiniBand or 10 Gigabit Ethernet connection, just like many 1G lines. At Wholesale Electric, Fife is using Xsigo's virtual I / O Director to separate processing, storage, and I / O. He said that by doing this, we have actually built our own cloud computing, because we can allocate processors, memory, hard drives and I / O as needed. Then, when you no longer need it, you can undo all the equipment and do other things. There are no strict restrictions on what we must do, and we are very flexible. Trend 2: Data and storage integration Joel Snyder, a senior partner of consulting firm Opus One and another member of Network World Lab Alliance, said that current data centers generally have completely separate data and storage networks. No one likes this situation very much. As long as people can merge these two networks together, that is exactly what they have to do. Snyder said, I think the hope is to remove the simple Fibre Channel and use Fibre Channel over Ethernet (FCoE). However, I still see people buying many Fibre Channels because they heard that this is the way to go, although our tests actually showed that the network is usually not a bottleneck. What you can do with Fibre Channel can also be done with 10 Gigabit Ethernet, and you will get the same or better performance, although this is not the idea of ​​storage LAN buyers and vendors. David Newman, president of Network Tes and member of Network World Lab Alliance, an independent testing company, said that this is the early stage of FCoE technology application, and many people are observing this technology. If there is nothing more, this technology has cost advantages. Newman said that in addition to the capital cost of equipment, there are also operating expenses. People running older Ethernets have less overhead than people running Fibre Channel. For economic reasons, configuring FCoE is cheaper than running a separate infrastructure. Today, both Brocade and Cisco have FCoE-enabled switches that fully support all optimizations and new mechanisms on Ethernet in order to provide Fibre Channel-style service levels. Other manufacturers are also entering this field. Therefore, it is currently possible to construct a viable end-to-end FCoE network to process data and storage, and at least equipment from the same manufacturer can be used. Interoperability has not been proven so far. Scott Engel, IT infrastructure manager at Transplace, a third-party logistics provider in Dallas, believes that FCoE is one of the two largest networks. Infrastructure changes will come to the company's data center in the next few years. The other largest network is 10 Gigabit Ethernet connected to the server. Newman said that the real mutation in the data center will occur within the next 12 to 18 months. At that time, 10 Gigabit Ethernet will replace Gigabit Ethernet on the server motherboard. That will have a variety of incidental effects, and achieving data center convergence is only one of them. Newman said, pay attention, this year will be the year when 40G switch port shipments reach "a considerable number". Wider network lines are required to accommodate higher speed server connections. Trend 3: Faster processors, greater integration So far, most large enterprises have stories of server consolidation around the theme of virtualization. They described the impressive ratio of physical servers to virtual servers, usually double-digit. However, some people say that data center integration is just beginning. IBM Power Systems manager Steve Sibley said that the level of maturity and comfort around virtualization is improving. This means that companies are showing willingness to provide more virtual machines on one system. He said that within this year, the Power 750 server will support 320 virtual machines on a single server. The Power 770 and Power 780 servers can support up to 640 virtual machines and plan to support up to 1000 virtual machines. Each physical server's ability to support a greater number of virtual machines certainly comes from the support of faster processors. In the case of IBM, the company recently introduced the 8-core processor Power7. This processor provides 8 times higher virtualization capabilities, scalability, and performance than previous generation processors. High-end Power 780 and 770 servers based on Power7 processors will be equipped with 64 Power7 cores. Intel is also preparing to launch an 8-core processor chip code-named "Nehalem-EX". This chip is expected to be launched in the middle of this year. Sibley said that if you start at the chip level, each processor core provides more powerful performance and the ability to add more than four times the number of cores on a chip will provide a lot of new capacity and capacity to put on a platform Adding more virtual servers will not significantly affect the performance or capabilities of the entire system. This design point is to let the system or product provide customers with more powerful integration capabilities, and integrate more virtual machines on a platform at a cheaper price than before. Trend 4: Infrastructure optimization Will your data center strategy one day include equipment for optimizing cooling and power supplies filled with a semi-tractable truck? Dan Kusnetzky, vice president of research at The 451 Group, a research company, said he can imagine this kind of thing, at least as an alternative to building new or expanding existing data centers. The road to software has failed. You may have to replace that truck with a new truck every three years. Doug Oathout, vice president of HP Converged Infrastructure, agrees with this view. He said that the concept of "Data Center in a box" is worth paying attention to. Enterprises are already using data centers similar to split pods or trailers outside their facilities to optimize server, storage, network, cooling, and power distribution resources for this containerized data center. Now, we see that performance optimization trends are entering the data center. Oathout said this is not to say that the data center will enter a parking lot full of semi-trailer trailers. At the time, companies without space, power, cooling, and capacity could accept the concept of a container data center and move that kind of similar assets to the data center. He said that we are not talking about the container itself, but this concept. We want to be able to say that I need 8 racks of servers, 4 racks of storage, 1 half rack of network equipment, and the power and cooling equipment it consumes, and optimize that way. Oathout said that it's much cheaper to combine data centers part by part than traditional risk-taking methods. Sometimes providing power and cooling is more efficient than transporting such equipment over long distances. Oathout said that building a data center with the largest capacity is a huge waste compared to building a data center as needed. Therefore, you can call this a retrofit trend. I will optimize what I already have, and use localized power, cooling, and energy to do the specific work I want to do in this environment. However, I take the next step, using multiple pods, entities, and components in the data center. It is unbelievable how efficient this is compared to building an overall data center. The overall data center with a megawatt power supply and 100,000 square feet of space cannot support the equipment you need for your next workload. (Compiled by Populus euphratica) Humidifier ,Home Humidifier,Tyre Humidifier,Humidifier Of Japanese Wine Bottle Ningbo Wellcome Trading Co., Ltd. , https://www.huike-homecare.com