In the keynote on day one of VMworld, VMware unleashed to the world EVO:RAIL. “What?” I hear you say, “but VMware is a software-only company!” Well, yes it is, and a software company is what it remains.
OK, so what exactly is EVO:RAIL? To give it its fancy Dan title, it is a hyperconverged infrastructure (HCIA). To take away all the gloss and marketing sparkle, it is nothing more than a reference architecture of closely integrated VMware products, wrapped up in a new HTML5 GUI that provides orchestration to make the magic happen. This is then run on a set of approved hardware from authorised partners. I know that Duncan Epping states on his VMware blog that it is not a reference architecture, and we can quibble over nuances and definitions if you wish. You could say it is a hardware appliance, but only from the position of the vendor providing the hardware.
Now, that might sound a tad harsh and belittling, but it is not really. Let’s think about this. As of today, EVO:RAIL’s partners consist of:
Interestingly, there is no HP and also no CISCO on the list. Is this by design, or was HCL approval too late? A question for another post, I think.
The partners listed above provide the hardware vessel for EVO:RAIL. It is at this point that it becomes an appliance. The partners, however, are provided a reference architecture to wrap around their hardware devices.
VMware has provided the following specifications:
An EVO:RAIL solution will, at a minimum, provide four nodes, and each node is to have the following:
- Two Intel E5-2620 v2 six-core CPUs
- 192 GB of memory
- One SLC SATADOM or SAS HDD as the ESXi™ boot device
- Three SAS 10K RPM 1.2TB HDD for the VMware Virtual SAN™ datastore
- One 400-GB MLC enterprise-grade SSD for read/write cache
- One Virtual SAN–certified passthrough disk controller
- Two 10-GbE NIC ports (configured for either 10GBASE-T or SFP+ connections)
- One 1-GbE IPMI port for remote (out-of-band) management.
Sounds like a reference architecture to me.
The resulting package, software, and hardware are sold as a single SKU, which means a single support point. And more interestingly, a single VMware support point. Yes, VMware is the point of contact for EVO support, including hardware support. I fully expect that the actual hardware support will be offloaded very quickly behind the veiled curtain, but from a customer perspective it is a one-call shop. This will make procurement an order of magnitude simpler.
One interesting thing I noted in the noise yesterday is that EVO will only be supported up to sixteen nodes: that is, 4 x 1 HCIA (remember, a single HCIA is four nodes). I find this a little odd, considering that it is based on vSphere and VSAN technology. Currently, VSAN can scale to thirty-two nodes, the full capacity of a vSphere cluster. But then again, this is just version 1.0, and I remember that the VSAN beta was at first limited to eight and then sixteen nodes before being released at thirty-two. I do not think that the four-node (sixteen per unit) limit is too restrictive a factor, and I fully expect that to increase to thirty-two quite quickly, perhaps even before the units appear on open sale.
- General-purpose applications
- Virtual private cloud (on-premises).
No big data applications or Tier 1 apps, like SAP, SQL, and BI. This does leave a gap for the other hyperconverged players, like Nutanix and SimpliVity.
So, what is my take on this?
To sum up, I think this is a very astute move by VMware. Hyperconverged appliances are the future. Marrying vSphere and VSAN and tightly coupling them in designed-for-purpose appliances shortens “time to usefulness.” As a 1.0 release, this shows significant promise. The introduction of an HTML5-based GUI is a boon: please migrate into vCenter client!
As to the future of EVO, I know that Chris Wolf (VMware’s CTO, Americas) has stated that it is not a competitor to VCE and that VCE still has a valid place. I personally think its position is open to change as EVO evolves and gains greater scale.
Share this Article:
Latest posts by Tom Howarth (see all)
- That Was the Year That Was: 2016 - January 16, 2017
- Docker Has Been in an Acquisitive Mood Again, This Time Pulling in Infinit - January 9, 2017
- Acquisitive LANDESK Bought Out - January 5, 2017