As a third-party reseller of used EMC and refurbished NetApp, we are receiving more and more questions about flash systems and drives. Flash can be a great addition to your environment and give you exceptional speed and capacity. But it’s still fairly costly to add, so you’ll want to be smart about where you use flash. It’s not necessary everywhere. Spinning disks can still give you what you need. This article will help you evaluate flash storage systems.
Flash storage systems use the ability to store memory cells in devices originating from floating-gate transistors. Traditional single-level cells store only one bit of information. Multi-level cell devices and triple-level cell devices, known as MLC and TLC respectively, can store more than just one bit by using multiple levels of charge to handle the floating-gate transistors. Flash storage appears to be a simple answer to application problems, but the choice of products, and where and how to implement it, can be multi-faceted.
Deciding the optimum solid-state storage to deploy involves critical analysis of the current storage environment you wish to enhance. It is essential to analyze the data flow, the bottlenecks, the workloads, and storage performance issues in the forms of performance problems.
It would be optimal if there was a magic genie that would analyze the issues in the storage scenario and type the perfect answer on the screen, but the variations on adaptability and real-world constraints make this unobtainable. However, good analysis of the variables will result in expert placement and utilization of the new products.
For flash storage to be the solution, the database bottleneck is the problem. Is the CPU slowing down, or is the issue in the host memory or the network bandwidth? Identify which resources slows by monitoring the utilization rates. If the CPU is churning to the max, then storage isn’t the problem, application architecture is. If CPU utilization is only at 50% or less, storage infrastructure is the issue. If this is the case, adding SSDs to the storage array could be the solution.
Adding SSDs into the storage system won’t help if the CPU is running at maximum. Instead, look to invest in a hybrid array or all-flash array. If the network bandwidth is restricted, discuss installing flash storage into the host server. SSDs are the most economical choice for expanding storage for the average situation.
PCIe cards offer more storage availability but also cost more per gigabyte. DIMMs can be connected onto the memory bus, making a much better solution than the previous use of DIMMs by connection to the memory board. This potential for server-side memory is just now being exploited. We expect many more advances in the near future.
Placing storage into the side-server reduces SAN traffic. Data that is accessed from the flash cache or flash tier doesn’t move the data across the network. This, in turn, frees up applications to support other servers.
Tiering is an option for flash storage that has implications for high speed storage in the data sets or database indexes. Tiering requires more flash capacity than caching does, as caching can employ a piece of software used to a specific server. Caching can be unpredictable in its performance and tiring to the lifespan of the storage, because of its write and erase capabilities. Determining the amount of cache storage needed will require data analysis on an on-going basis. Write-around caching should be thoroughly explored as some write-caching methods can corrupt data.
Questions about how flash drives or systems fit into your SAN storage environment? Contact us. Feel free to comment below or contact us directly at Sales@Reliant-Technology.com.