Today finally marks the announcement of the general availability of VMware VSAN. This is VMware’s most public beta to date.
So, What Is VSAN?
VSAN is true integrated software-defined storage. It is embedded into the vKernel, it is a distributed object store, and as such, it is resource efficient. There is no requirement for an extra virtual machine on each host.
During the launch, VMware CEO Pat Gelsinger stated, “The hypervisor opens up new opportunities…and is the Goldilocks zone…it can see the guest and it can see the legs.” This is true, and it is the deceptive simplicity of the product that beguiles. Installation is simple and quick. Is it a Ronseal product, i.e., does it do exactly what it says on the tin? Yes, but it is a Hokey-Pokey product, too. It scales in, it scales out, it scales up, and it scales down.
During the design of the product, Pat stated that they followed the basic rule of virtualisation: the use of abstraction.
I cannot get over how slick this product is, and I can say that from the position of actually using it during the beta period.
Pat remarked that the value proposition of the product was as follows:
- Radically simple
- High performance—local to compute
- Dramatically lower TCO.
I can agree on points one and two, and can give a reticent agreement on the third point, because we still have no real understanding of how this product will be licensed. Will it be licensed per node, or will it be licensed by capacity as per the Nexenta model? And more to the point, whatever the licensing model, how much will it cost?
But back to the technical specs. As alluded to earlier, this has been VMware’s most public beta ever. The vast majority of spectators expected that VSAN would GA with the ability to grow to sixteen nodes, but VMware CTO Ben Fathi dropped the bombshell that it will actually scale to thirty-two nodes on GA. That is awesome and quite a significant figure, as thirty-two nodes also happens to be the maximum size of a VMware cluster. What also amazed was the capacity and performance that can be squeezed out of the product:
- 32 hosts
- 3200 VMs
- 2 million IOPS, and
- 4.4 petabytes.
Also impressive for a 1.0 release is the seamless integration with other VMware product sets, like SRM and View.
Flash devices are used not just to optimise performance but to also provide redundancy by utilising devices in other hosts for a second write copy, which is then slowly written to the underlying spinning disks.
So to sum up, do I think this is a game changer? Yes! Do I think that traditional storage will need to worry? Most likely; the savings in hardware costs and ongoing support have the potential to be mind-boggling. The fact that the implementation is so simple, coupled with the fact that they have a large gamut of vendors already on board to provide ready configured nodes, is impressive. All that said, there is still one thing I would really really like to know: how much will it cost? However, somehow I think I will have to wait until the tenth of March, when I can actually purchase it to find out, because without knowing costs, how can you actually define your true value proposition, above and beyond pure marketing hype?