Hyperconverged technology is an exciting development in scale-out computing. Read about a hyperconverged product that is pitched as best suited for IT departments with zero to five people.
Hyperconverged data centers have a single node that contains storage, networking, compute, and virtualization. Generally, it's scalable by just adding more nodes as you go.
Scale Computing's HC3
At a recent Storage Field Day I got to see Scale Computing present its hyperconverged product, HC3. The HC3 node comes in three flavors: the HC3 (which grows to 100 VMs), the HC3x (which grows to 200 VMs), and the HC4000 (which can grow to 400 VMs). The HC4000 is in a limited introduction right now; currently, it can scale up to 8 nodes.
Jason Collier, a founder of Scale and its CTO, spoke first and said something that rings true from what I know of the company: "simplicity in everything we do." Scale designs solutions that target data centers for SMBs (Figure A).
We were shown a demo of the user interface, which reflects Scale's keep-it-simple core tenant. There are tabs for each layer of the hyperconverged node, including Summary, Virtualization, Storage, Replication, Snapshots, System, and Hardware.
It's easy to do upgrades; as long as there is room on the nodes, the Scale nodes will complete a rolling upgrade by live migrating VMs to upgrade one node at a time with no downtime.
Most of the popular operating systems are supported as guests (recent Windows servers, popular Linux distros, etc.). There is also built-in redundancy for the VMs. Scale compares it to being a bit like a RAID 10 solution. Though you may lose some usable storage due to this redundancy, most scale-out architectures are built with this kind of redundancy. It also protects against disk failure and node (or server) failure.
There's no need to go in and carve up storage and specify things like iSCSI targets or HBAs etc. -- every VM has access to the entire storage pool. Generally, each node has four or eight drives, and the drive size depends on your needs.
There is currently no flash storage or tiering done within the nodes, though Scale is looking at that, which I think is probably necessary to be competitive in the hyperconverged market, even if you are targeting SMBs.
Price and support
The entry-level list price for essentially a 3-node cluster is about $25,000. When you consider what you'd have to spend on servers, storage, and then a virtualization layer on top of that, I think it's a decent price. The price includes one year of premium support for free (premium is the only support they offer), so this is 24x7x365.
As an engineer who works for a value-added reseller (VAR) that specializes in data centers, I've seen the trend moving towards using managed services, and Scale appears to have positioned itself quite nicely.
It seems like an increasing number of SMBs and even enterprise companies are hiring fewer people as full-time admins. In this day and age of automation and cloud, it's possible to have a really lean IT department. With Scale, you can literally rack and stack the HC3 nodes, cable it, and have it up and running in a couple hours. In his presentation, Jason Colliers reminded us that Scale is best suited for IT departments with zero to five people. When he says zero people, that's exactly what he means.
It's an interesting time for these hyperconverged technologies, and I'm really excited about this kind of scale-out computing. Are you? Let us know in the discussion.