Server Fitness Program: How to Improve server capacity

Server performance is becoming more robust as CPUs, memory, and storage technology evolve. For example, CPU manufacturer Tilera Corp. recently released the TILE64 family multi-core processor. The processor contains 64 separate processor cores, and each core is a fully functional processor. Each core contains L1 and L2 caching, which means that these cores can support a full operating system (OS) independently. The physical server technology has elevated memory management to a new level. For example, the HP Proliant DL580 G7 Server, which has 64 DIMM slots and can support up to 2TB of capacity DDR3 memory.

However, why do many data centers still bother with server capacity planning, and how to solve the problem of server selection and hardware resource planning?

The key to solve the problem is to understand the server hardware resources planning, deployment and management misunderstanding, to avoid such errors.

Understanding Server Capacity Planning

Any server on-line, the most important part is the planning stage. Assuming that there are many data center implementations, it is critical to choose the server technology that will be used. In many cases, IT administrators will be throwing money at the most expensive and state-of-the-art systems they can buy. While this may work well in the initial phase, this is not the most stable solution.

Industry experts say planning is the most important, preferably in the application environment before entering the development link to determine.

"When we talk about capacity planning, almost everything is dependent on the data center infrastructure," said Timothy O ' BRIEN,MTM Technologies Inc. system Advisor. "Before deploying any servers, we have to set a reasonable expected baseline and identify the services that each server will run." ”

The following two major issues need to be taken into account when analyzing server capabilities:

Understand the business that the server environment will host in the future. For example, what is the main purpose of implementing system virtualization? Or do you need to use a separate physical server?

Increasing numbers of users, services, and applications will put pressure on computing resource requirements. How will the environment expand upgrades in the future? In short, are you ready for future application expansion?

A bit of server capacity planning is often overly valued, and that's end-user performance. As the IT administrators have said, if end users are unhappy with performance issues, the server pilot will soon end. For example, a strong and essential system in a storage area network (SAN) environment. However, because you have a lot of storage, it means that you have the desired performance. If the SAN is overallocated, the user will feel the performance degradation of the application service. One of the basic methods to guarantee the user experience is to select a tool that can perform numerical test and behavior tracking. Any unexpected performance fluctuations can be traced to a change operation, a hardware failure, or a problem that occurs in other data centers.

The server does not need to spend a lot of money to keep it running steadily. Having a thorough deployment plan to understand the tasks that the server will run can save a great deal of overhead.

Understanding server Types and resources

Before the server is deployed to the formal environment, agents must understand what the server will be used for. By analyzing its workloads, IT administrators can reasonably allocate resources and determine server models and sizes. Server hardware resources can easily be exhausted and experienced IT professionals will tell you that resources are actually very limited. Even servers with sufficient computational capacity may run out of resources due to workloads and underutilized applications.

"Server model and resource management have always been a challenge for data center engineers." Understand the extent to which the workload will be available, and decide how much resources to provide, "O ' Brien added." "For example, a simple Web server for displaying an internal Web site does not require multi-core processors or large amounts of memory. However, more computing resources are required to run SQL Servers with multiple user connections at the same time. ”

Pre-planning can reduce the pressure on resource allocation. Knowing the load that the server will run and planning ahead allows the environment to have enough resources to handle and meet the requirements. Because almost all server infrastructures are not immutable, engineers must be prepared to face possible environmental changes at all times.

The most important part is that, whenever you have to remember, the server does not just finish installing. Choose the right resources for the server to host future work challenges, while the load and the computing resources that are required are always variables. So, planning the server in the future will be how to run, need to put in the right memory, computing power and storage space.

When embarking on resource allocation planning, the following three key points need to be noted:

All configurations are not immutable. Tuning server resources is a normal job. Some resources can be assigned to a running server in real time.

Monitor the server environment. It is important to understand the current server resource usage. Load management involves monitoring server overload and discovering peak demand, allowing engineers to allocate resources more efficiently.

Learn about applications, operating systems, and platforms. Never assume that an application or an operating system will run in a rut. With Service update packages, new user and environment changes, applications may consume more memory, storage, or even CPU resources at any time.

Virtualization

Planning a server as a virtualized platform is somewhat similar to planning a separate physical server, but there are many differences. Engineers need to evaluate which servers will run on this platform and what workload will be generated. As previously mentioned, it is similar to the need to plan ahead before deploying a physical server:

What are these servers designed to run?

What load will the server carry in the future?

Does the load, operating system, platform, database, application requirements increase with the business? Have we taken these growth into account, and this physical server can be full of corresponding requirements?

By understanding the programs that the server will run, engineers can make server configuration planning and cost more reasonable. What applications will run on a physical server can be discussed and changed at the planning stage. Whether this server will run Citrix XenServer or VMware or install SQL Servers in the virtual machine, the trick is to understand what kind of application the allocated hardware resources will host. With this approach, the ability to plan for future development will be more effective and accurate.

Server Capacity Planning Best practices

Planning and establishing a server environment relationship mapping can help engineers better resource allocation optimization for load. Resources are often exhausted by overallocations, so knowing the load they will run can save a lot of time and reduce later problems before the server is online.

There are a number of powerful tools that can help us measure requirements and determine how the server will run. A tool called Up.time, developed by Uptime software, can help administrators monitor servers, virtual machines, cloud services, hosting services, and other applications. Using Up.time's graphical server monitoring software, the administrator can analyze all the key service resources in the data center through graphs, and it supports various operating system platforms. In more depth, you can refine information such as CPU, memory, disk, process, load, network, user, service status, and configuration data so that moderator engineers understand how they will optimize the server.

When you are planning your server capacity, follow these points:

Any perfect deployment needs to go through the testing process. In a test environment, you can adjust the server and make it clear how to configure to get the best results. Take the time to understand what applications the server will run and how much impact these applications will have on the overall environment.

Learn about the applications that the server will run. Learn about an application, a database, or an operating system that consumes resources, avoiding insufficient or overallocated resource allocations. Resource-sensitive workloads include EXCHANGE,SQL databases and other database applications. Low resource consumption loads are similar to authorized servers, single service applications, or simple Web servers.

Be prepared to respond to changes in the environment. Being prepared and aware of the resources available to optimize workloads can help improve resource management capabilities and responsiveness, shorten time to change, and make server infrastructure more balanced and stable.

Never invest money on a server just to fix one or two problems. Plan budget expenses for the servers that need to be purchased and analyze them for reasonableness. It is common to replace or upgrade stand-alone components during the hardware lifecycle. However, just as the price of server technology has been changing, it is sometimes more cost-effective to buy new servers directly than to upgrade CPUs, hard drives, and even memory.

As server technology continues to evolve, IT administrators will devote more effort to the planning process, deciding on server capacity and resource planning. Whether they are managed entirely by physical servers or environments containing virtual machines, or by maintaining a database server farm and understanding their role in the data center, they can guarantee their stability in the present and future.

About the author: Bill Kleyman,mba,mism, a super tech expert in the field of network infrastructure management. His engineering tasks include large-scale virtualization application deployments and business network design and implementation. He is currently the technical director of the World Wide fittings Global Manufacturing Company, which has offices in China, Europe and the United States.