Teach you how to diagnose the performance of the server

Data center managers should learn to interpret the symptoms of performance problems and determine what factors are causing the problem. The slow response of the server makes everyone crazy — — end users are anxiously waiting for Microsoft PowerPoint files to open, senior executives hope to receive emails immediately, as data center managers, you must listen to and solve their complaints these questions. In today's increasingly virtualized environment, server performance issues are more common and more serious than ever before. Erwin Vollering, director of virtualization services at consulting firm GlassHouse Technologies, said: "In a physical environment, each server has its own SCSI controller, disk, memory, and processor components. Now, you run 10 to 15 servers in a shared environment on a physical host. These servers handle the same tasks at the same time, use the same resources, and crash once the physical host fails. ” So, how do you go about diagnosing these problems in order to solve the problem and continue to resume operation? Here are a few tips to help you get started. Make sure the problem is the server
Jesse Rothstein, CEO and founder of ExtraHop, says that many of the performance issues that seem to come from the server level are often caused by problems at the network, database, and storage systems. Rothstein said: "In the end, you need to consider and analyze these issues in general, but it is actually tricky. “When there is a delay, we need to see if this delay first appears on the server or at the network or client level,” said Dan Klimke, Global Product Marketing Manager at Fluke Networks. Tracking and analyzing symptoms takes a lot of time. & rdquo; IT departments often connect multiple servers to provide resources (such as applications) to end users. Klimke said: "You may have a web-based front-end server that talks to the application server, and the application server talks to the database server, all of which are connected through some kind of network. So the first step in troubleshooting is to make sure the problem really comes from a server, not from the network. & rdquo; Getting cumbersome applications
It's entirely possible that an application causes the server to react slowly or simply shut down. Klimke said: "The application may be too sloppy & rsquo; (chatty) & mdash; & mdash; that is, such applications are written without using large frames with large amounts of data, sending fewer frames, and The smallest frame is used, so it is only necessary to send a large amount of data back and forth frequently to complete the task that a larger frame can accomplish; in contrast, using a large frame is much faster and more efficient. ” For example, a command from an application might be sent to the server, causing problems with the server. Klimke said: "Look at the actual data traffic entering the server, you can find out which link the packet entered, and what command was sent to the server before the server was down. This necessitates a view of the actual data traffic on the network connection, as well as the ability to interpret past commands and frames that were sent before the server was down. &b;Clean up dirty networks
Rothstein said that dirty networks are networks where packets are lost, reordered, or fragmented, which can drag down server performance and cause server performance degradation and server load. Big. He explained: "Most people don't realize that network protocol stacks like the TCP stack have both fast and slow paths. The fast path is efficient, optimized for the packets it is to transmit, and the next packet can be transmitted in order. But if you transmit packets out of order, or inadvertently transfer packets from Layer 4 down to Layer 3, the server and operating system will consume a lot of resources to reassemble those IP fragments, or reorder or Reassemble the TCP fragment. & rdquo; Virtualization may complicate this issue. Rothstein said: "Your hypervisor is a scheduler …… share processor time across all running virtual machines and guest operating systems; if you switch from one virtual machine to another, it will inevitably appear delay. If the configuration is wrong, the network packet will be delayed, so that it is discarded by the underlying network stack; once you reach the tipping point, performance degradation is not only obvious, but also has catastrophic consequences. &b;Using Monitoring Tools
Rothstein said that no matter whether it is an automated third-party tool that performs SNMP polling to record server health indicators, or a simple free open source tool, you should make full use of monitoring tools. Vollering said that if you are using a virtualized server, monitoring tools are especially important. He said that VMware, Microsoft and Citrix, the three major virtualization technology providers, provide excellent monitoring tools. He said: "These tools can tell you what is putting pressure on your environment, whether it's processor usage, memory usage or disk usage. " Vollering said that even if those tools are available and deployed, many companies often don't know if they can use them or don't know how to use them. In addition, these tools bring flood-like information and alerts. He said: "You may have to spend a whole day looking at the various charts. & rdquo; So, how can we make the most effective use of these tools? Vollering said: “Read the manual carefully and configure the tools for your environment. That way, if an alert occurs, you know that you need to explore the alert, not because the tool is still configured with default thresholds and is flooded. &b; Best Tips: Checking the Server's Physical Resources
CDW's Internal Solutions Architect Kelly Sather said that once you've identified a performance issue from the server level, you should double-check it. The physical resources of the server. He said: "The best case is as simple as adding another processor or adding more memory. "Dan Klimke, Global Product Marketing Manager at Fluke Networks, added: "It's important to look at whether your processor utilization is too high for many processes. We found that many problems were caused entirely by the disk being full. &b;Basic Tips: Make sure the server is up to work
Sather says that because your hardware is compatible with newly installed or upgraded applications, it doesn't mean the application will run smoothly. He said: "The resources in your server need to be commensurate with the load it is going to handle. For example, someone migrated to Exchange 2010 and moved out of the old version of Exchange running smoothly on a dual-processor dual-core system with 24GB of RAM. But because Exchange 2010 is a more resource-intensive system, they later discovered that the system could not run on old hardware. & nbsp; Two other tips: Correcting and adjusting the load
The load from multiple virtual servers can far exceed the processing power of the physical server. Erwin Vollering, director of virtualization services at consulting firm GlassHouse Technologies, said: "You need to know your limitations." If you take a snapshot while processing monthly financial consolidation numbers, your storage system will not be able to handle all of those transactions at the same time. & Rdquo;

live migration of virtual machines inside
CDW's solutions architect Kelly Sather said that with the increased workload, the surge in the number of virtual machines, you may need to find a cause The problem virtual machine is migrated to another host, or the "normal" virtual machine is offloaded to another host to make room for virtual machines that need resources. VMware's vMotion and Microsoft's Hyper-V (and other technologies) allow you to migrate virtual machines to other hosts in real time, without the need to connect to end users.