StorReduce forms a critical part of your cloud storage infrastructure. You need it to be reliable and resilient in the face of unexpected outages and failures - wherever they might occur. StorReduce has always been designed to protect the integrity of your data, while still operating at unparalleled speeds, but until now a failure in the physical infrastructure underpinning the StorReduce appliance could temporarily cut access to your data.
This is no longer a problem, as StorReduce now supports highly available deployments via our new clustering mechanism!
StorReduce’s clustering support is based on etcd, which leverages the robust raft consensus algorithm to co-ordinate failover during physical hardware failures and network partitions.
You can craft deployments of StorReduce to span availability zones and even regions to ensure you can maintain access to your data even in the unlikely event of a complete failure of a cloud datacentre.
In the past clustering solutions have been clunky and somewhat complicated to configure. We really focused on making our solution “just work” in a seamless and intuitive way. A new cluster can be created with the click of a button:
And to join a new server to the cluster it’s as simple as copying and pasting some text:
At its core the clustering mechanism uses a master-slave configuration. The cluster elects a master server to process reads and writes, and should that master server fail for any reason the cluster will quickly and seamlessly promote a slave to become the new master. If the cluster can not agree on which node should become the new master (e.g. because more than half the nodes in the cluster have failed) then all writes will be halted to protect the integrity of your data.
This is just the first cut, in future iterations we plan to support enabling read access on slave servers, to help balance load across the cluster, as well as other more advanced features.
Clustering is available in version 2.8.0+. Click here to request a trial of StorReduce.