IT managers are being convinced by storage vendors that solid-state flash storage will accelerate their applications by magnitudes. But after deploying solid-state storage, many see only modest performance gains or no performance improvement at all. So, what happened to the promise of faster flash performance for applications?
There are several reasons why this can happen outside of inflated product claims (this is another discussion in itself). First, solid-state storage can expose additional bottlenecks in an infrastructure (server, network, hypervisor, applications, etc.). When storage is no longer the bottleneck, other flash performance problems may become noticeable. Historically, servers, operating systems, applications and networks have been designed to communicate with the speed of hard disk drives (HDDs). When solid-state is deployed, the rest of the infrastructure may need to be retuned to support the faster speed and response time.
For example, say a new all-flash array is installed to accelerate performance for virtual desktops. But, the storage area network was designed with enough network performance to handle the performance of HDD arrays. Replacing those HDD arrays with all-flash arrays where the performance is two to 10 times faster with latencies in the microseconds (instead of HDD milliseconds) can suddenly add an unexpected load to the network — especially if that network is a few years old. Also, the HBAs and I/O subsystem in the server may be overloaded. Both of these things can have a major impact on overall flash performance.