Data replication and self-healing architecture

 

Automatic live data replication is inherent within Joviam’s cloud architecture, without further user configuration. Blocks backing virtual disks are automatically distributed and replicated across the cluster upon creation, utilising dual redundant 40Gbit QDR InfiniBand (IB).

In contrast, server outages at traditional cloud providers have resulted in hours to days of downtime, with resulting business disruption and data loss from asyncronous backups that may or may not have captured in-flight transactions between the outage window and the last backup.

Within the Joviam platform, for every write to disk:

• The write is sent across the IB framework and written to multiple nodes.
• The storage driver waits for confirmation from all nodes that the blocks of data have been written.
• The storage driver then delivers confirmation of the write to the virtual machines.

Node availability within the cluster is constantly monitored by our Software Defined Storage (SDS) system to ensure prescribed levels of data replication are consistently maintained.

Reliability = Stability

If our SDS detects that a node is either having issues or is offline, data replication of the missing blocks to other nodes within the cluster begins immediately, thereby maintaining prescribed levels of data replication. When the original node is recovered, any changes made to storage are automatically updated and replicated – this entire process is entirely automatic, requiring no user involvement.

Live data replication vs. geo-redundant backups

There are three main advantages to automatic live data replication over geo-redundant backups:

1. Synchronous data replication, whereby data is written to two different nodes at the same time. In a geo-redundant scenario, data is written to a main node first and sent afterwards to a backup node in a different location. As a result, any new data created during this transfer disappears resulting in data loss.
2. No additional setup required by the user. Typically, the onus is placed on the user to architect data redundancy and backups themselves. With Joviam, data replication is inbuilt within our cloud architecture, occurs automatically and is completely transparent to the end user.
3. Automatic AND immediate node failover. In the event of a node failure, most cloud providers require the user to deploy their own backups, thereby tying downtime duration solely upon user reaction time. Joviam’s self-healing architecture sees automatic and immediate node failover, without the need for backup deployment by the user.