I’ve been witness to a number of converged infrastructure systems at the large corporate and enterprise scale environments. These solutions are typically quite robust in nature, requiring a number of disciplines to implement and maintain. You’ll most likely need a background in a variety of areas, such as blades, SAN, virtualization, and route / switch networking in order to be successful with a deployment. A larger corporation or enterprise scale environment will have these skill sets on staff, or have a means to provide them on an ad-hoc basis (via consultants or contract employment).

Smaller companies, however, are typically left out in the cold when it comes to leveraging integrated compute infrastructures. They usually employ a small number of IT staff who are asked to perform a wide variety of tasks, often labeled a “jack of all trades, master of none.” And indeed, IT generalists are the oil that make a lot of SMB entities run well; I’ve been there myself. There simply isn’t the time and budget to get the training and exposure to the equipment necessary to build a big stack of gear in a lot of cases. And speaking of budget, converged infrastructure usually isn’t cheap by any means. Just the networking piece, be that 4 or 8 Gb SAN or 10Gb ethernet, can blow out a SMB’s budget.

This is why I find Scale Computing‘s HC3 offering so interesting. It’s purpose built to target the SMB entity. As per the company’s own site:

HC3 was built for medium-sized organizations with small IT staffs. Our products are widely used in the healthcare, education, manufacturing, financial services, and government industries. They have been specifically built for easy deployment and to be managed by IT administrators who want to spend more time focused on business needs and applications, rather than troubleshooting infrastructure and hardware issues. Ideal for main offices, remote offices and DR sites, HC3 fits any type of environment where applications need high availability, not high costs and complexity.

When I was first introduced to this idea, I immediately drew some comparisons to other offerings in the marketplace today (such as Nutanix). I think the point that Scale Computing is trying to make is that the idea of “share nothing” infrastructure isn’t itself new, but the way they’ve put a spin on it is, along with their price points and target market, is something that is needed today.

Under The Hood

Scale Computing has been shipping product since 2009 in the form of their unified scale out NAS/SAN nodes in the form of S-Series and M-Series systems. Again, their target audience is those who don’t have much of a SAN or NAS background; people who just want a product that works without having to deal with the complexity of HOW it works.

With the launch of the HC3 product, the company is advancing their current storage node system into a fully integrated compute solution. The underlying hypervisor, while not being from VMware, will have a feature set in parity along the VMware Essentials Plus licensing level, and will start in a four node layout, scaleable up to eight nodes. This is a self imposed limitation, as the company plans to grow beyond eight nodes in the future as they advance the product.

Scale Computing had no trouble figuring out how to stripe the storage layer across nodes, as their current offerings already do this. In the diagram below, you’ll see that physical disks are presented as virtual disks and/or shares for file serving. Their secret sauce, the ICOS Services, provide all of the back end lifting for the nodes to provide services such as NFS, CIFS, iSCSI, and more.

The storage layer is already tried and true from the S-Series and M-Series clusters

From a user perspective, very little of this is actually presented in a way that needs to be managed. They see a single pool of compute and storage that can be carved up and utilized as desired. Again, keeping in line with the idea of being “super simple to use” means taking care of the complex pieces behind the curtain. Virtual workloads can be migrated from one node to another, and in the event of a node failure, surviving nodes will start up the workloads.

In the event of a failed node, workloads are restarted elsewhere on the cluster

The amount of virtual workloads that can be run simultaneously depend on the number of hardware nodes that are included in the cluster. Scale Computing has published a few numbers around nodes to VM counts, ranging from 30~50 virtual machines on a four node system, to 60~100 virtual machines on an eight node system.

User Interface and Features

I had the opportunity to work with the user interface on an HC3 system, and must say the HTML5 based presentation layer is really slick. Most of what you need is represented as simple gas gauges and “filled cups” type of charts. Below is a graphic with the main virtualization dashboard and a dialog box to create a new virtual machine.

The use of HTML5 is very slick

From a Disaster Recovery perspective, at launch you can do replication from cluster to cluster on a per volume basis. This mainly goes back to the days of the nodes being a storage product, and replication was only concerned with volumes. Moving forward, the idea on the road map is to offer replication on a per VM basis.

Support and Upgrades

Scale Computing offers full support over both the hardware and software using their own internal staff in Indiana via their “ScaleCare Support” packages. They offer a two hour SLA phone response for any issue you determine is critical, with one year of support bundled into the purchase of a cluster. The option to add further support (such as a 3 year bundle) also exist.

One down side that I’m not particularly excited about is that upgrades currently require a maintenance window for an outage. The update process does give an estimated time of outage to update the cluster, but still, a node architecture shouldn’t have downtime. However, I have been assured that the team is working towards that goal, and that in the future, a node by node rolling upgrade will be achievable.

I’ll also point out that if you currently own their storage node products, the company is allowing you to upgrade to the HC3 solution via a firmware upgrade.

Thoughts

This sounds like a really good solution for those who want a simple, converged infrastructure solution in the SMB space. The company has already developed a history for offering a solid product in the converged storage node market, so many fears over a “new player” are mostly set aside. While I am purposely not going over pricing numbers (that’s not my bag), I will share that they have some very low price points for the amount of simplicity baked into their solution that should make it an easy win against a “roll your own” type of deployment. If you’re at VMworld, drop by their booth and take a look, it is a really neat product to watch in action.

Additionally, Stephen Foskett will be hosting a Roundtable discussion with Scale Computing at VMworld, including myself and other past Tech Field Day delegates, that should provide you with some additional informative goodies.