IBM TotalStorage DS6000 Series: Performance Monitoring and Tuning

476 IBM TotalStorage DS6000 Series: Performance Monitoring and Tuning

Goals of benchmarking

Today, customers have to face difficult choices regarding the number of different storage

vendors and their product portfolios. Performance information provided by storage vendors

could be generic and often not representative of real customer environments. To help in

making decisions, benchmarking is a way to get an accurate representation of the storage

product’s performance in simulated application environments. The main objective of

benchmarking is to identify performance capabilities of a specific production environment,

compare performance of two or more storage systems. Including the use of real production

data in the benchmark may be ideal.

To conduct a benchmark, you need a solid understanding of all parts of your environment.

This understanding includes not only the storage system requirements but also the SAN

infrastructure, the server environments, and the applications. Recreating a representative

emulation of the environment, including actual applications and data, along with user

simulation provides efficient and accurate analysis of the performance of the storage system

tested. The characteristic of a performance benchmark test is that results must be

reproducable to validate the tests’ integrity.

Benchmark key indicators

The popularity of a benchmark is based on how meaningful its workload is representative.

Three key indicators can be used out of the benchmark results in order to evaluate the

performance of the storage system.

򐂰 Performance results in a real application environment

򐂰 Reliability

򐂰 Total cost ownership (TCO)

Performance is not the only component that should be considered as benchmark results.

Reliability and cost effectiveness are also parameters that must be considered. Balancing

benchmark performance results with reliability functionalities and total cost ownership of the

storage system will give you a global view of the storage product value.

To help customer understanding of intrinsic storage product values in the marketplace,

vendor-neutral independent organizations developed several generic benchmarks. The two

most famous organizations are:

򐂰 SPC - Storage Performance Council: http://www.storageperformance.org

򐂰 TPC - Transaction Processing Performance Council: http://www.tpc.org

The popularity of such benchmarks is dependent on how meaningful the workload is versus

the main and new workloads which companies are deploying today. If the generic benchmark

workloads are representative of your production, you can use the different benchmarks

results to identify the product you should implement in your production environment. But, if

the generic benchmark definition is not representative or does not include your requirements

or restrictions, running a dedicated benchmark designed to be representative of your

workload will give you the ability to choose the right storage system.

Requirements for a benchmark

You need to carefully review your requirements before you set up a storage benchmark and

use these requirements to develop a detailed but reasonable benchmark specification and

time frame. Furthermore, you need to clearly identify the objective of the benchmark with all

the participants and precisely define the success criteria of the results.