Flash and hybrid storage are the latest nirvana technologies, as they promise to relieve all storage performance problems. Determining which applications justify the need for flash and exactly how much flash storage to deploy remains the biggest questions. The answers usually revolve around understanding application workload performance requirements and having the data for a realistic cost/benefit analysis for solid state deployment. Here are five ways to help you stay within budget as you evaluate upgrading to flash storage.
1. Understand The Workload I/O Proﬁles
Storage architects need to gather intelligence about the unique characteristics of their application workloads in a given environment. The concept is to extract and analyse statistics on production workloads from the infrastructure to establish an I/O baseline profile. From there, you can project I/O growth trends. By capturing all of the attributes of the production environment, highly accurate workload models can be created which enable storage managers to stress test storage product offerings using THEIR speciﬁc workloads.
2. Create Workload Models
Workload models represent storage traffic as it enters the storage array. The profile will represent random vs. sequential data, reads vs. writes, distributions of blocks sizes, distributions of files sizes, queue depths, data content types, access patterns, and other key metrics that comprise the storage traffic.
3. Generate Workloads
Once these workloads are executed against the storage system(s), you can determine the I/O performance characteristics and limitations of any given storage platform. This allows storage planners to size their deployments without over or under provisioning. It also enables the projection of when more resources, like network bandwidth, storage IOPs, etc., will be needed to maintain SLAs and whether or how much ﬂash storage will be cost-effective.
4. Inline Deduplication & Compression
Unfortunately, these technologies impact application performance – in some cases by a factor of 2X or more. You need to fully understand their performance impact before making the purchase decision. Every storage vendor has a different implementation and uses different algorithms. Accurately measuring the performance of deduplication and compression depends on generating data content patterns sufficient to stress a storage array. Ensure your load generation appliance can generate compressed and deduped content.
5. Easy To Use Reporting & Analysis Tools
This means being able to easily understand and compare IOPS, latency and throughput across a variety of your specific application workloads. The ability to automate the testing and report generation process, without any custom scripting, is critical.
Performance validation appliances enable infrastructure planners to bring more automation into the storage performance planning and validation process. These purpose-built appliances can be used to help organisations pre-determine if flash storage is truly justified and test all configurations prior to making production deployment decisions.