Guest Blog from Product Manager Gil Nadel at Infinidat
Infinidat Sets a New Bar for vVols Scale
Infinidat has been a VMware partner for almost a decade, with a long list of VMware integrations and capabilities that make life easier for virtual infrastructure admins building software-defined data centers on top of our InfiniBox storage systems. Our recent release of InfiniBox 6.0 software added support for VMware Virtual Volumes (vVols) with two key attributes in mind: ease of use and scale. This article will discuss the latter and walks through how we worked with the VMware Cloud Solutions Lab to demonstrate a new bar for vVols scalability – three times larger than the previous largest vVols design.
When I write “scale” I mean not just the petabyte-plus InfiniBox capacity configurations we typically ship, but also the underlying system design parameters that support massive workload consolidation – things like the numbers of volumes/vVols, number of snapshots, and number of pools (read: tenants/datastores). We designed our vVols implementation to match the other massively scalable InfiniBox capabilities I describe below.
First, let’s talk about total capacity. Our typical customers measure their InfiniBox systems in petabytes, and our biggest arrays scale to beyond 10 petabytes of effective capacity per system today. Further, the Infinidat software architecture is designed for 10s of exabytes of theoretical capacity, so as the needs of Infinidat customers grow and the economics of different media types become more feasible, we will introduce storage arrays with matching capacity.
What about datastore size? A vVols datastore is backed by a Storage Container, and those are implemented using InfiniBox pools. InfiniBox pools consolidate capacity and QoS management for a group of filesystems or volumes and provide storage admins the flexibility to control capacity utilization. If desired, a pool can be created to use the entire capacity of the InfiniBox storage array, so a single datastore can grow to tens of exabytes.
Based on our flexible pool approach, InfiniBox can provide up to 1,000 vVols datastores per system, which is 60x our largest-scale competitor – and that’s only a testing limit, could be increased via RPQ.
To complete the picture: a single InfiniBox volume or vVol can theoretically be 259KiB, approximately 500 PiB. This is much more than the current VMware virtual disk size limit of 64 terabytes, but when VMware increases the supported capacity metrics, InfiniBox is ready.
vVols Scalability in the Real World
Let’s take this back to earth. Imagine you set up your vCenter and vSphere servers with InfiniBox backend storage – maybe you’re already using InfiniBox with VMFS datastores. There may be thousands of virtual machines running, with one or more virtual disks per VM, let’s say 5,000 VMs in total. Suppose you wanted to migrate those to a vVols datastore on InfiniBox, so that you can benefit from native integration into advanced InfiniBox data services, such as zero-penalty snapshots that you can create and delete instantly. Assuming on average 1.5 virtual disks per VM, then each VM would require 3.5 vVols: 1 config vVol, 1 swap vVol (when the VM is running), and 1.5 vVols for virtual disks. That’s a total of 17,500 vVols for 5,000 VMs. And imagine you wanted snapshots for these VMs as well, say 10 snapshots on average. The snapshots would require 15 vVols per VM, an additional 75,000 vVols just for the snapshots.
You end up with 92,500 vVols supporting 5,000 VMs… can InfiniBox support that many? Short answer: yes, and more!
When we designed the vVols implementation, that was one of the anchor scenarios we had in mind. To prove the scalability, we ran a test in the VMware Cloud Solutions Lab:
- We used an InfiniBox F6240 storage system running software release 6.0.0, which provides 1.4PB usable capacity.
- We used a vSphere cluster with 8 servers, each connected to redundant Fibre Channel fabrics.
- We created a VMFS datastore for a template VM, and we mapped the vSphere cluster to the Protocol Endpoint.
- We mounted the VMFS datastore and created a template VM running Linux, with 20 virtual disks.
- We also mounted the vVols datastore.
- We then cloned the template to create 1,000 VMs on the vVols datastore and took 5 snapshots of each VM.
- We next powered up all the VMs and took 4 more snapshots of each VM.
The result was over 200,000 vVols on the InfiniBox storage array, more than 3X the previous largest-scale solution.
Here are some action shots from the lab environment:
1000 VMs in vCenter
A sample VM setup
And finally, InfiniBox shows over 200,000 vVols!
Many of our customers are already working at that scale with traditional VMware storage approaches, and we are happy to help them and other large-scale VMware shops transform with our new vVols implementation. For example, one of our joint partners Mark III Systems was able to get a preview of our new capabilities and shared their excitement during our launch. Their president Stan Wysocki told us that as customers scale out their VMware environments, they aim to gain operational efficiencies and reduce the risk and cost of their virtual infrastructure. By implementing InfiniBox with VMware vVols, “We are able to accelerate each client’s time to value, while we prepare and equip them for the needs of next-generation VMware and modern app stack environments.”
I couldn’t have said it better myself. Learn more about Infinidat and vVols here.
Viva Las vVols!