Application virtualization on the server

As the use of terminal virtualization and application virtualization increases, IT managers must realize that server requirements may change significantly because the virtualization technology selected for application delivery is different.

Traditional non-virtualized application delivery requires applications to be installed on a physical server, and end users can access applications through a LAN. Since this delivery model only consumes 5% to 10% of the server's computing resources, few people will pay attention to the overall configuration of the server.

As long as the server can meet the computing resource requirements required by the application, administrators will pay little attention to the resource consumption of the application delivery until there are obvious performance problems related to the overall server configuration. For example, a large number of users may reduce network performance, and the Administrator may bind multiple NICs to solve this problem.

Terminal virtualization applications have completely changed this practice. Virtualization improves server usage and enables physical servers to process more concurrent tasks. However, administrators are faced with the challenge of planning. For an application delivery server, sufficient computing resources must be allocated to support end users or terminal nodes. To avoid resource depletion and insufficient performance, continuous management is required. In addition, the virtualized server must be elastic enough to minimize downtime and user interruption. Virtualization has higher requirements for IT experts to coordinate and balance resources.

Application virtualization Server

The application virtualization server allows users to access independent instances of an application installed on the server. However, application virtualization requires high I/O performance. For example, the selection of storage locations may have a great impact on application virtualization.

"At any time, when deploying a [storage Region network], the first thing to consider is the I/O problem," Lan ParkerThomson Reuters, senior website service administrator of Global Information Resources Company) said. "We have been paying close attention to the development of flash drives recently ."

Disk I/O performance is very important for storage arrays, But I/O problems may also come from the network. For example, an Ethernet-based SAN, such as iSCSI or FCoE, may cause network bandwidth bottlenecks. Therefore, configuring storage on an independent application server can simplify potential network congestion problems. Network I/O performance can also respond to the bandwidth-intensive application stream. The bottleneck occurs when the application is delivered to the terminal as needed, this is different from running an application on a central server.

Parker said that memory is not a significant problem for application virtualization, because the 64-bit operating system server can easily support hundreds of GB or even terabytes of memory capacity.

VDI servers

Virtual desktop interface (VDI) The server hosts the entire desktop instance on the central server. Only user input and audio/video output information are exchanged through the simple terminal device of the user side, this type of terminal is generally called "thin client" or "Zero client ". Servers hosting the entire desktop instance are more sensitive to resource overhead than application virtualization. Therefore, you should consider upgrading to a more powerful CPU, memory, and storage I/O. Once the desktop instance is loaded and started running, the network I/O requirements are reduced to a very low level.

Local Storage can effectively improve VDI performance, but SAN is a more popular storage platform because it also provides a single point of management function.

"If you are planning to deploy VDI on the SAN, the bandwidth will become very important," Parker said. "The next step is the memory and Disk Throughput, because you will find a lot of page exchanges and other activities ."

The VDI instances on the delivery server will continue to grow, and the disk subsystem must meet the increasing performance requirements. For example, Parker finds that when the VDI instance starts the system, the primary activity on the primary storage is to write the disk, and the disk array method to improve the write performance should be used, the RAID 5 array with poor write capability may affect the performance of the storage subsystem.

Personal desktop instances can also improve performance by adding the number of server CPU cores. Therefore, selecting a CPU with multiple cores during server configuration can improve the server's VDI hosting capability.

"The larger The number of CPU cores, The less The core, The faster it is," said Bob Plankers, The Lone Sysadmin technical consultant and blog columnist.

Consider the single resource requirements of applications

Virtualization applications may have additional requirements on some server configurations. For example, medical imaging applications designed to process large volumes of files may have high requirements for memory and storage I/O. Peak Performance may occur after application virtualization and when services are delivered to multiple users. For example, the Stratusphere tool of Liquidware Labs Inc. Can help administrators diagnose the resource overhead required for applications before migration to VDI deployment.

Similarly, consider the growth of other virtualization applications, such as computer-aided image and rendering tools. Integrated visual technology, similar to RemoteFX in Windows Server 2008 R2 SP1 and 'hdx in Citrix Systems Inc. s XenDesktop 4, allows you to use powerful graphics cards on terminal servers for computing. This is another step forward in application delivery, but it administrators need to weigh the impact of such technological advances on the server infrastructure. "What type of blade server can support mainstream PCIe interface graphics cards, but also extra space and Cabinet occupation issues," Parker said.

However, there is no niche requirement for major commercial applications at present, and virtualization applications only have some influence on servers more or less. "This is what we care about, but it does not have much impact on how we purchase servers," Parker said.

　Elastic delivery of clusters and Applications

Important servers that provide enterprise users' key application delivery usually require sufficient flexibility to ensure availability. The server itself provides some elastic functions, such as onboard RAID controllers used to manage local disk storage and redundant power supplies. Traditional server cluster integration tools or recently developed Virtual Machine redundancy technology can protect the load from unplanned downtime. For application virtualization, this is a step that requires more attention, because any application terminal will affect a large number of users.