Scale Computing is known for their mid-market storage solutions that offer customers the ability to "scale" their systems without storage specialists, massive overhead, or huge upfront investment. By having a linear growth model, Scale allows you to right-size the storage system to exactly your needs whether it’s archiving, VDI, or a virtualized server environment.
Scale offers two models, "S" and "M" series to give their customers a choice of platform and performance. The "S" series clusters in sets of 3 minimum, comes equipped with dual 1GbE ports, SATA drives, 3 different sizing options ranging from 1TB to 4TB usable per cluster member, all in 1U per node. The "M" series provides a workhorse and a wide array of storage and speed options, including 10GbE and SAS drives in that same 1U form factor. Both systems run ICOS (Intelligent Clustered Operating System) and support iSCSI, NFS, and CIFS plus the added bonus of a highly reliable, redundant, and secure storage platform. You can read up more on Scale’s storage options here if interested.
Let’s be honest though, cheap storage is nothing new. While Scale has a great (slightly unique) offering, the market is overflowing with vendors who’ve got the "next big thing" and it’s hard to sometimes not get lost in the chaff. That’s why they’re introducing something that makes perfect sense (and validates their name).
Enter HC3. Scale is gearing up to enter the virtualization world, getting ready to shake up the status quo and do something that a few others have tried, but with a major twist. This is a different approach to the datacenter. What they’re doing is taking those powerful storage systems and embedding a hypervisor, right inside. Now we’ve all seen vBlock and FlexPod, and the prices that go along with them.
When you’re talking about a pre-built rack with SAN, networking, and servers all built into it with the licensing models of those particular companies… it adds up fast. This isn’t a datacenter in a rack solution, this is an expanding datacenter in a box. An entirely converged infrastructure. It’s a model we’ve seen before; build the hardware around being an all inclusive solution. The difference is that Scale is simply taking the systems they already have, utilizing some of that extra horsepower and giving you access to it in the form of a fully manageable KVM environment that clusters. There is no blade cluster within the box, the storage controller becomes your compute environment. On top of that, the management is simple to use and the price is right for any small/medium business out there trying to move away from the one-off servers but not break the bank.
This evening I sat down with the guys at Scale Computing and some members of the Tech Field Day community to see how it all works and ask some questions. Here’s the breakdown.
First we’ll start with the beginning. The idea behind the new offering came from a previous venture in 2005 utilizing Xen on a clustered filesystem in an HPC environment. It never really got off the ground as it was a purpose built system for a niche market and was put on the back burner for a while. After a couple years, the company was sold off, but the idea and the tech stayed with the founders until they formed Scale Computing (a bit of foreshadowing in the name there). Since 2009, Scale has been successfully shipping a storage platform that they built in house, gaining a customer base and seemingly spending the past 3 years getting ready for today. Now that they have a solid foundation to build upon the idea is back in the spotlight and ready to ship.
I must say, at first glance it looks good.
The underlying OS is an RPM based system, similar to RedHat 6.3 and utilizes KVM. When asked why KVM over other hypervisors the response was that KVM links into the kernel and interfaces with the filesystem better than any other system out there. It’s a mature platform and has a lot of support in the community so it wasn’t hard to build with. KVM required very little modification to run smoothly with the chosen filesystem (the changes are contributed back to the opensource community). KVM is a well respected and feature rich hypervisor in the opensource community and doesn’t get a lot of attention from the enterprise market.
Note: The system could work with any hypervisor that supports libvirt (the virtualization API), which potentially leaves the door open for quite a few others, but as of right now there are no plans to expand into other virtualization platforms. The biggest deterrent for moving to a major hypervisor over KVM is licensing, this model works well and can be sold at its price point because there is no licensing required.
They chose GPFS because of the reliability, scalability, and performance. There were comparisons done to other clustering systems (mainly GlusterFS and Lustre™) but IBM’s solution proved to be the best fit.
Upon initial release the system is aimed to feel very similar to VMware Essentials and will have the features below:
- VM import
- Live migration/fault tolerance
- Replication between clusters
- Thin provisioning
- Crash consistent snapshots
- Lightweight clustered management via HTML5 web portal
The road mapped features (that I could be told) hope to bring it to the level of VMware Enterprise in the near future and include:
- SDN integration
- Per VM replication
- Future models may be developed to build out more memory/CPU resources as needed and tailor your environment
What licensing? This is a value added feature. If you buy the M series storage, it’s included. Yup. Even better, if you are currently using Scale as the storage for your VMware environment, fear not, these systems can coexist. This not only allows you to test their system out, but makes migration away from VMware quite easy.
This system is specifically designed to fill a hole in the small/medium business market and that’s where Scale wants to focus. An IT shop with a couple of overworked guys who don’t have time to babysit an entire virtual environment as we know it today will find this to be ideal. It frees up management time and simplifies the process of deploying and maintaining a virtual infrastructure. There may be expansion in features and usability, but it was made clear to me that there are no plans to grow the systems out to enterprise level and leave the smaller shops behind.
This is something to keep an eye on. After reviewing the info and asking questions, the only down side I can see to this whole system is that adoption of KVM may be difficult to push. Most companies won’t see it as an issue until a software vendor pushing their wares doesn’t recognize it. As one of the other TFD delegates pointed out, if it runs Windows it’ll work. Unfortunately, "Our software is designed to run on VMware or Hyper-V" has been a consistent theme throughout a recent RFP process I went through. Running Scale’s converged systems may cause support issues if I were to choose them. This may be less of an issue in the near future though. Redhat is currently on a push to get KVM recognized and certified by vendors out there. Riding on the coat tails of Redhat has never really been a bad place to be for anyone and could prove to be a driving factor in adoption. Other than that I can’t see any other problems.
However, I do see a few different scenarios that could play out here.
Outcome 1: The system fails. Not what anyone wants to see happen (maybe someone does, but not me), especially with cool tech like this.
Outcome 2: Flash in the pan. Performs well enough, sells well enough, but never shakes anything up and disappears into the background noise.
Outcome 3: The system is fantastic and exceeds expectations, gaining traction in higher education, doctor’s offices, and branch offices. I don’t think we’d be looking at just another product that offers virtualization in this situation; we would be have a product that potentially becomes the catalyst SMB/SME needs in the virtualization world. For many years the major players out there have catered to large enterprise and offered a low cost, stripped down (barely affordable and still overly complicated) version to those that can’t afford to invest heavily to get what they need. This might be the kick those providers need to step up and sell a competitive well rounded small market solution. (Yes, I’m a jaded engineer)
I have high hopes for Scale’s HC3 system and would love to see something like this rock the boat, but only time will tell.
*Note: My trip to VMworld 2012 was partially funded by Scale Computing via the Tech Field Day organization. I was given early access to some of the above information and I am under no obligation or expected to promote their product, discredit a competitor, or even write this article.