A brief overview of the new Virtual SAN 6.1

captain-vsanWith the announcements just made at VMworld 2015, the embargo on Virtual SAN 6.1 has now been lifted, so we can now discuss publicly some of the new features and functionality. Virtual SAN is VMware’s software-defined solution for Hyper-Converged Infrastructure (HCI). For the last number of months, I’ve been heavily involved in preparing for the Virtual SAN 6.1 launch. What follows is a brief description of what I find to be the most interesting and exciting of the upcoming features in Virtual SAN 6.1. Later on, I will follow-up with more in-depth blog posts on the new features and functionality.

Stretched Cluster

The most significant enhancement, in my opinion, is support for VSAN stretched cluster. By using the fault domain concept introduced in Virtual SAN 6.0, customers can now protect virtual machine across sites with their VSAN 6.1 cluster. The smarts now included in VSAN 6.1 will ensure that one copy of the data is placed on one site, a second copy of the data is placed on another site, and that the witness components are always placed on a third site (witness site). And the witness site does not need to have a physical ESXi host; we have created a special witness appliance, which is basically an ESXi in a VM, for holding the witness components. In conjunction with vSphere HA, DRS, and VM/Host Affinity rules, you can run your VSAN stretched cluster in an active/active configuration with running virtual machines on both sites (if you wish). In the example below, there is a 2+2+1 deployment; 2 nodes on the preferred site, 2 nodes on the secondary site, and 1 witness node.

7. FD view on completionHowever, just like vMSC (vSphere Metro Storage Cluster), there are a number of considerations to keep in mind when deploying a stretched cluster. One needs to consider the bandwidth and latency between the data sites, as well as the bandwidth and latency to the witness site. Of course, there is also the network topology to consider; should I go with a stretched L2 (switched) network or a L3 (routed) network? We have created a VSAN Stretched Cluster Guide which will take you though all of these design considerations, installation steps and failure testing scenarios. Anyone considering stretched VSAN should read this first. I’ll follow-up with a number of addition posts which delve in the details of the above considerations soon.

2-node VSAN support

Virtual SAN 6.1 also introduces support for the 2-node cluster. Well, it is actually 3 nodes, but one of the nodes can now be a witness appliance, the same witness appliance mentioned in the stretched cluster case. So in fact, you will have a 3 node VSAN cluster, but two of the nodes are physical ESXi hosts and the third node, which once again only hosts the witness components, will be an ESXi host running in a VM. This will be ideal for those ROBO (Remote Office, Branch Office) use cases. In the screenshot below, there is a 1+1+1 environment, but of course for ROBO, both the preferred and secondary host would most likely be located in the same office. The witness could be located at either the remote office, or back in the main DC. Lots of choices for customer’s here to suit their needs.

1. 1+1+1Health Check v2 Updates

 Health Check for VSAN has been a huge success with our customers. So much so, that in vSphere 6.0u1, health check is now automatically included with both the vCenter Server 6.0u1 and ESXi 6.0u1. Now there is no need for customers to manually install RPMs, MSI or VIBs. These are already present.

There are some other enhancements too of course. There is a new suite of health checks for stretched cluster, but you’ll only see them if you deploy a stretched cluster VSAN.

There is also alerting/alarming in vCenter based on health check test failures, which is a very sought after feature. And of course, you can associate actions with these alarms like you could any other alarms, and send SNMP traps or emails.

cluster-summary-alarmsNew flash technology support

nvme-ultra-dimmAs if the all-flash VSAN announced with version 6.0 wasn’t fast enough, VSAN 6.1 has support for new flash technologies. These include Intel NVMe and Diablo Ultra DIMM. I saw a demo of the Diablo Ultra DIMM earlier this year, and that thing flew! A great enhancement in my opinion, and its great to see these new storage technologies in Virtual SAN.

Multi-Processor Fault Tolerance (SMP-FT) support – Virtual SAN 6.1 now supports vSphere Fault Tolerance. Unlike vSphere HA, which needs to restart a virtual machine in the event of a failure, this feature provides continuous virtual machine availability with zero downtime. This can be an extremely useful feature for mission critical applications that wish to run on VSAN, as it will avoid any outage in the event of hardware failure.

That’s it in a nutshell. As I said in the intro, this is very much a brief introduction. I’ll get into a lot more detail once the VMworld madness is over.

11 comments
  1. Hi Cormac,

    Great post as always.

    Will it be possible to share the witness for the 2 node ROBO configuration or will each VSAN cluster need a dedicated witness VM running in the central site?

    Many thanks as always
    Mark

Comments are closed.