kantver - Fotolia

Five IT troubleshooting tips to beat storage, network bottlenecks

When something goes wrong in the data center, there are many possible culprits, ranging from the network to storage. Here are some tips to identify and solve common bottlenecks.

When problems arise in the data center -- and they will -- there are plenty of options when it comes to the first step toward solving them. Whether that's filing through logs for a change in procedure, checking the hardware for damage or working through a network bottleneck, the answer is out there.

To rectify any issue that's costing you time and your company money, know the ins and outs of your IT infrastructure. Combing through servers and storage systems -- either on premises or in the cloud -- can point you in the right direction.

Here are five IT troubleshooting tips for the data center.

Identify server issues

As a server struggles to keep up with the many demands thrust upon it, issues can arise. The first step when IT troubleshooting for servers is to determine how widespread the issue is. To do this, identify commonalities between user complaints to gauge the extent of the problem. Inspect hardware and software -- and collaborate with co-workers -- to sift through potential issues.

One of the simplest and most overlooked IT troubleshooting options is checking the logs. Look at Microsoft Windows Event Viewer logs or syslogs to identify recent changes in the server, and help narrow the search for irregularities. If all else fails, call the vendor for assistance to get another set of eyes to scan for an issue.

Identify CPU bottlenecks in I/O processes

There are multiple IT troubleshooting techniques to address network bottlenecks, ranging from quick and inexpensive repairs to long-term, high-cost overhauls.

In a CPU-limited mainframe, the available I/O processes significantly affect workload performance. Batch processes take the brunt of the bottleneck, as they compete with higher-priority workloads, such as online transactions, for the CPU processing time. As a result, batch jobs slip to the bottom of the dispatcher chain. Even after other I/O processes are completed, the rise of the batch job up the dispatcher chain can hurt performance.

With online transactions in an I/O process in a Customer Information Control System or Information Management System, a CPU bottleneck slows it down in a similar fashion to a batch process. To prevent this bottleneck, tap into in-memory data, such as buffer pools or reference tables, wherever possible.

Consider SDN when using solid-state storage

While solid-state storage can increase application performance, its high IOPS can create network bottlenecks in the data center. As workload and storage performance increase, the network's inability to handle the increased traffic volume results in a bottleneck.

Consider a software-defined network to help combat this issue. Software-defined networks respond more programmatically to traffic changes and demands than a manually configured network, helping to reduce bottlenecks. Data deduplication is also an option, as it reduces storage demands. Organizations should perform deduplication at the source system, rather than the target or destination system, to reduce bandwidth consumption.

Explore new data flows, containers

There are multiple IT troubleshooting techniques to address network bottlenecks, ranging from quick and inexpensive repairs to long-term, high-cost overhauls. Changing the data flow in a network is an example of a low-cost fix that can keep latency low and reduce backbone traffic. To do this, you can colocate storage nodes across multiple servers -- an approach Google has taken.

Containers can also help reduce traffic, but every server instance needs to run the same operating system. The startup time is fast, but similar problems with bottlenecking can occur if the apps are I/O-intensive.

As for the higher-capital options, inter-switch connections and networking are on the horizon, with new hardware and software capable of multilane and high Gigabit Ethernet rates.

Think about server and storage location

The farther away data is stored from a server, the more time it takes the server to access it. Intra-rack local storage reduces this time and network traffic across the data center. However, there are temperature concerns in a shared-rack environment, since hard disk drives require a cooler temperature than servers. This complicates shared rack design and management. In cutting-edge operations, solid-state drives in the data center eliminate this issue because of their ability to work at a higher temperature.

Local storage often meets most needs, with a network backup in place to serve as a fail-safe in the event of a server crash. A virtual area storage network setup would work well, but the servers need to have storage space for data, which can drive up costs of the servers.

Next Steps

Should syslogs be mandatory for log management?

How unpredictable workloads impact mainframe capacity

Identify and prevent virtual server bottlenecks

Dig Deeper on Data center capacity planning