Most will know Atlantis from the days of ILIO Persistent VDI and ILIO Diskless VDI. I would expect that most even consider it a niche product with the single goal of accelerating desktops. I have to admit, I was shocked to find that this is no longer the case. It is well known that I have been quite vocal in my thoughts about this company. People I know and trust have had bad things to say about its performance and stability. Admittedly, this is hearsay and also very old knowledge. But mud sticks. I also have to say I have undergone a road to Damascus moment regarding this company and its products. They have obviously improved on stability; further, the performance and functionality have gone stellar.
I think I need to offer a little background on my historic position on Atlantis. I have been aware of this company since startup. I remember Chetan standing all lonely in his booth at VMworld 2008, explaining to anybody and everybody who would stop what this wonderful new technology would do. I admit that, although it intrigued me at the time, I was at a loss to see a real world application. Fast forward eighteen months: I was talking to some very good friends of mine about a project they were involved with, and up popped Atlantis and ILIO again. It was safe to say that their comments about the product were not complimentary, so back into the box marked “Emmm” Atlantis and ILIO went.
Jumping to the present day, I was recently briefed on the company’s latest offerings. After broaching the subject with the great and good, all I can say is colour me impressed. It seems that it is much more than an IOPS generator.
The company has recently announced its ILIO USX product. This moves it into the server arena, arguably a much larger pond in which to go fishing.
What is USX? What does it do? How does it do it?
At the 30K ft. level, USX is an accelerator that effectively offsets storage performance from the SAN/NAS or DAS level and moves it to RAM speed. No, this is not a reference to the blunt instrument at the front of a Roman galleon, but rather to good old-fashioned DRAM in a server. The product is, however, aimed at the server market, even though Atlantis’s traditional focus has been on VDI.
Using patent-pending technology, the product compresses, de-dupes, coalesces, thin provisions, and does IP processing of VMs (not just vSphere-based environments).
How does USX do this? At its most basic level, it utilizes RAM in the hosting server to run the IOPS by making a storage tier in RAM. Technology similar to VMware’s Changed Block Tracking is utilized to move the useful blocks of storage into the effective RAMDisk. If this were all the product did, it would be useful from an IOPS and latency point of view; however, it also de-dupes the data inline to further remove duplicated data to increase the consolidation ratio of machines in the RAMDisk.
This results in a very high level of improvement for consolidation and performance. According to Atlantis, up to five times more VMs than usual can run on your current installed storage base. USX does this by unlocking the latent storage overhead incumbent in the legacy storage tier—all those RAID 1-0 volumes created for performance can be utilised for data, thus driving up your ROI on your storage.
Atlantis as a company is engaging in several interesting partnerships with companies like NetApp, VMware, and Dell to sell bundled solutions. This is very astute on Atlantis’s part, as those names are trusted advisors to their client bases. Now, Atlantis has visibility to that list, and its partners’ sales teams will be acting as stealth employees; thus, Atlantis derives value from both products.
To sum up, from a technical standpoint, the product is most definitely very impressive—and very disruptive. It complements legacy storage by uncoupling performance from the Tin, thereby allowing the release of the latent storage overhead previously required for IOPS performance. It addresses a demand for performance and capacity with minimal changes to infrastructure.
It will be interesting to see how quickly support for other hypervisors comes on stream. This is due to Atlantis’s architecture, by which it utilises a VM on each host to manage the Atlantis layer. This is probably less of a technical issue than others presented by similar products in the same space, but what is key is the pricing strategy. Today it is vague, and perhaps that is understandable, given that it is a new product. A software-based solution should be much easier to sell and to deploy quickly than traditional Tin. However, without clear pricing, this isn’t going to happen. The pricing needs to be clear and concise; otherwise, the value proposition will be hard to unearth and sales cycles will remain long.