Deploying Stretch Clusters to Increase Resiliency

Published On: 1st October 2018//2.3 min read//

You may have seen our blog from a couple of weeks ago on the importance of creating a highly available environment, to avoid downtime when faced with hurricanes and large-scale natural disasters. However, the majority of physical disruption and disasters that organizations face are localized to a single, specific site. Minimizing or eliminating localized outages is easily achieved by effective deployment of stretch clusters.

StorMagic stretch clusters can be deployed in a single location (in the same rack, room or building) to prevent against a localized incident that can take a server down (like water damage, electrical outage or human error). However, the technology also provides the option to create a cluster that stretches up to 1,000km – which is plenty of distance to protect against incidents happening across campus or even a city or region. SvSAN creates a highly available environment through synchronous mirroring and a remote witness node, such that if one side of the cluster suffers an unexpected outage – all applications running at both locations keep on running. When service is restored, the data stores are automatically rebuilt and HA is restored within minutes.

With most other competitive solutions on the market, deploying a stretch cluster isn’t so simple, as many face challenges with bandwidth and latency. Sometimes there is too much data, trying to travel too far, at the same time. In many cases, those who are smart have decided against deploying a stretch cluster and those who have risked it have ended up with a low performance system that makes them want to tear their hair out.

HOWEVER, StorMagic SvSAN is famously lightweight, which naturally extends to SvSAN deployed as a stretch cluster. You can read a full summary of the latency testing findings in our stretch cluster white paper, but in short we found that SvSAN as a two node stretch cluster showed constant response times and IOPS up to a 30ms latency. For peak performance, it is recommended that latency does not exceed 10ms, as to reduce the possibility of unexpected behaviour. This can be up to 1,000km for our solution.

The amount of bandwidth required varies based on the amount of data that needs to be passed between the mirror. We recommend the bandwidth should be a minimum of 1Gb between SvSAN nodes.

The testing showed that SvSAN as a stretch cluster can work incredibly effectively with high latencies and therefore over long distances, making it an ideal method of creating resilient storage, while still enjoying the simplicity, cost effectiveness and flexibility that you get with SvSAN.You can read more about our lightweight software-defined storage solution, SvSAN, here. Or if you would like to find out how we can meet your specific requirements, you can contact the team at [email protected]

Share This Post, Choose Your Platform!

Recent Blog Posts
Go to Top