Should I Set Benchmarks for My Storage Environment?
Yes, infrastructure benchmarks are an important metric to develop, understand, and maintain in an enterprise environment. Benchmarks are key tools to determine application-supporting infrastructure design and components, and they provide a stable baseline for root-cause analysis activities. Benchmarking for storage...
June 7, 2007
Yes, infrastructure benchmarks are an important metric to develop, understand, and maintain in an enterprise environment.
Benchmarks are key tools to determine application-supporting infrastructure design and components, and they provide a stable baseline for root-cause analysis activities. Benchmarking for storage can be more complex, but also more important, than for other infrastructure components, since the data path for large enterprise applications is typically comprised of multiple systems, each with its own contributions to performance, bandwidth, and latency.
Having a repeatable process to create storage environment benchmarks also enables you to maintain a successful tiered-storage environment.
The following steps should help you to successfully create a baseline storage environment benchmark:
1. First, understand and map out your server and storage infrastructure from HBAs to arrays.
2. Create a map of how data is allocated per application and the associated data paths.
3. Create a list of components, including operating system-based logical volumes, host bus adapters, Fibre Channel switches, and arrays. The storage administrator must understand the base configurations and how they affect performance, capacity, and response time.
4. Create a process to load test and gather metrics for each component in the data path.
The baseline benchmark should help to determine the best performance of the storage environment, regardless of server, network, or application issues.
As an example, using a standard single server, Fibre Channel, and monolithic array architecture to create a benchmark, the administrator would define processes to gather metrics from each layer, as listed below:
Server
Logical volume IOPS
Logical volume response time
Disk queue length
Fibre Channel Switch
Port throughput
ISL utilization
Port utilization
Array
Front end adapter throughput and utilization
Cache utilization
Back end adapter throughput and utilization
Disk drive IOPS and utilization
These metrics will define the process for creating the benchmark.
The next step is to determine an industry-standard-based load-testing tool or script to determine the maximum performance of the architected solution.
Once this is done, the baseline can be used to define where application storage should be allocated within a tiered storage environment. It also can serve for root cause analysis and troubleshooting for application performance issues.
Tim Arland, Principal Consultant for Storage Solutions, Forsythe Solutions Group Inc.
Read more about:
2007You May Also Like