Flash storage is all the rage these days and there’s an important aspect of flash deployments that comes up often in discussions in my neck of the woods: have you considered the impact of flash on your network?
To dig into this topic a bit deeper, I sat down with Kyle Roblyer, OEM Marketing Director at Brocade.
There will be more on this topic so be sure to click ‘Follow’ at the top right of the page and stay tuned!
Kyle, most all-flash conversations in the press are about storage. What things should a customer consider when designing a network to support high-speed, flash-based workloads?
The excitement of flash has rightly been focused on the benefits of storage and application acceleration. Due to its massive performance benefits, flash can eliminate some of the typical bottlenecks in the environment but it also has the potential to simply move those bottlenecks into other infrastructure components like the network. For flash and all flash environments, a predictable, resilient and always available network design is required. One way to ensure that your network is prepared for an all flash deployment is to leverage Brocade's free SAN Health check tool, which helps you identify potential problem areas within your network.
Check out Brocade's free SAN Health check tool to be sure you're network is ready for Flash. You can download SAN Health check tool here.
Another factor in all flash performance is latency. Latency is often talked about like a storage only issue. Is it?
Latency problems can be caused by just about any component in the system. Sometimes slow storage devices are the bottleneck, sometimes it’s the server not being able to process I/O requests fast enough. Other causes include oversubscription in the network, or even damaged cables or SFPs. If we look at storage networks, Fibre Channel has long been the standard for mission critical storage environments, due to its performance, scalability and low latency characteristics. Because of its widespread deployment, and established position as the default low latency storage networking protocol, Fibre Channel networks actually become a key enabler in realizing the full potential of flash and all flash storage systems. Based on this, we expect most storage customers to continue to use FC as the protocol of choice for flash environments.
Do converged solutions have different issues when it comes to latency and delivering high IOPS?
Converged platforms such as Hitachi’s Unified Compute Platform (UCP) play a big role in flash deployments. In fact, there are over 40 Hitachi UCP solutions that include Brocade standard FC and IP rack switches as well as embedded Compute Blade switches, all of which have been architected to meet the requirements of an all flash environment. The converged stacks are pre-configured by Hitachi to meet the rigorous latency and IOPs performance requirements for various application environments.
Because managing all these discrete components can become an administrative nightmare, Brocade integrates the network intelligence captured from its Fabric Vision tools into Hitachi’s management software, providing a single pane of glass to simplify overall management of the UCP stack.
So … what’s on the horizon?
Because flash storage behaves differently than physical spinning disk, the industry has been forced to take a look at new standards and advancements in how the network handles this type of storage. For example, NVMe over Fabrics is an interesting space to watch. This emerging specification aims to take the performance and latency advantages of Flash and extend those benefits over Fabrics. Scott Shimomura, Senior Director of Product Marketing at Brocade, recently wrote on this topic. If you're new to the topic of nVME, here's a primer.
If you want to learn more about networking requirements for flash environments, here’s a paper written by George Crump.