Synchronous replication to ensure RPO (RecoveryPointObjective) arrives on the Nutanix AHV hypervisor with version 5.17 of the AOS operating system.
With the release of AOS version 5.17, Nutanix responds to a request expressed by a large number of customers. Synchronous replication arrives in AHV (in addition to asynchronous and near sync replication), which therefore allows disaster recovery, with unplanned failover to begin with. “This was a major request from our customers for their critical applications where you cannot lose data,” explained Christophe Bardy, EMEA solution strategist at Nutanix. “Until then, all clients using synchronous replication were under ESXi or Hyper-V. Today, this function is also available with our hypervisor. (…) They trusted us on AHV with Flow for security, and they needed synchronous replication”. Synchronous replication is the first brick of Metro AHV, a feature announced at the last .Next in Copenhagen. . This is the first Metro AHV brick, announced at the last .Next in Copenhagen. Synchronous replication was already available on the VMware hypervisor, but unlike the metro cluster on ESXi which works by datastore, Nutanix brings granularity to the VM with AOS 5.17. Note also that near sync replication on all hypervisors – for 100 VMs – goes from 60 to 20 seconds, which allows a snapshot three times more frequent.
Main technical condition to ensure this synchronous replication: a latency of less than 5 ms between the two remote sites to ensure a perfect copy of the data and ensure a zero shortcut RPO. “As with our competitors, synchronous replication requires a latency of less than 5 ms or risk significantly degrading write performance,” explains Mr. Bardy. It’s both a question of distance and hardware to ensure this minimum latency between two metro datacenters with Equinix for example. The second brick of Metro AHV is expected soon and will ensure a live migration from cluster to cluster in order to ensure planned switches: change network or evacuate all the VMs to a second datacenter to carry out a maintenance operation. Testing of failover scenarios will also be possible without having to shut down VMs.
A third brick will provide the option of a witness to automate failover operations in the event of a site failure. A witness node generally ensures consistency between the two sites when the two instances (active and passive) can no longer communicate with each other. “Some customers do not want disaster recovery in manual mode, but automatic and a witness is necessary for the automatic failover. However, it will be an option and the customer will be free to use it or not,” explained the technical manager of Nutanix. This third brick will be available for all hypervisors supported by AOS, namely AHV, ESX and Hyper-V.
Another feature that ships with AOS 5.17, Foundation Central allows a cluster to be configured remotely after it has been racked and connected to the network. This function was used a lot during confinement with very large customers to deploy servers in agencies or stores, assures Christophe Bardy. Foundation Central is integrated with Prism Central. The latter, which makes it possible to manage several clusters with control of all the modules, sees its capacities reinforced: the limit has been raised to 300 clusters (with a maximum of 32 machines per cluster) and 25,000 VMs in a single batch.
.Next next September
With this version of AOS, still billed per processor core, Nutanix upgrades its in-house hypervisor to offer all possible combinations and effectively compete with ESX for all customers operating complex topologies. All the bricks are not yet available, but the next .Next Experience conference, expected in September 2020 in virtual mode, should make it possible to come back to these pre-announcements.