It’s been a little while since I talked about Atlantis Computing and things have developed a bit since then. They’ve added a bunch of new features to USX, including, amongst other things:

• Atlantis Stretched Cluster;
• Volume-Level Snapshots; and
• Atlantis USX SmartSnap.

I was recently lucky enough to have the opportunity to be briefed on their latest developments by Priyadarshi Prasad, Senior Director of Product Management at Atlantis Computing.

HyperScale CX-4

Atlantis Computing recently announced a new addition to their HyperScale range of products – the CX-4. If you’re familiar with the existing HyperScale line-up, you’ll realise that this is aimed at the smaller end of the market. Atlantis have stated that “[t]he CX-4 appliance is a two-node hyperconverged integrated system with compute, all-flash storage, networking and virtualisation designed for remote offices, branch offices (ROBO) and “micro” data centres”.

Atlantis Computing have previously leveraged Cisco, HP, Lenovo and SuperMicro for their hardware offerings and this has continued with the CX-4. The SuperMicro specs are as follows:

Dell FX2

Atlantis also let me know that “Dell is teaming with Atlantis to provide the entire line of Atlantis HyperScale all-flash hyperconverged appliances on their PowerEdge FX2 platform. Atlantis HyperScale CX-4, CX-12 and CX-24 appliances are now available on Dell servers through Dell distributors and channel partners in the U.S., Europe and Middle East, shipped directly to customers”. Here’s an artist’s interpretation of the FX2.

As far as the CX-4 goes, the Dell differences are as follows:

• Form factor – 2U 2N or 2U 4N
• Memory per Node – 256GB – 768GB
• Redundant Integrated 10GbE switch

Resiliency

Resiliency for the cluster comes by way of a mirror relationship between the two nodes in the CX-4 appliance. Atlantis also provides the ability to define an external tie-breaker virtual machine (VM). In keeping with the ROBO theme, this can be run at a central site, and multiple data centres / appliances can use the same tie-breaker VM. There is also high availability logic in the CX-4 system itself.

The tie-breaker is ostensibly there to keep in contact with the nodes and understand whether they’re up or not. In the event of a split-brain scenario, there is a fight for the tie-breaker (a single token). But what happens if the tie-breaker VM is unavailable (e.g. the WAN link is down)? There’s also an internal tie-breaker operating between the nodes, handled by a service VM on each node.

Simplicity and Scale

One of the key focus areas for Atlantis has been on simplicity, and they’ve gone to great lengths to build a solution and supporting framework ensuring that the deployment, operation and support of these appliances is simple. There’s a single point of support (Atlantis), network connectivity is straightforward, you can have IP configuration done at the factory, and everything can be managed either centrally via USX Manager or individually if required.

The CX-4 can be used as a gateway to the CX-12 if you like, simply by adding another CX-4 (2 nodes). Or you can choose to scale out, depending on your particular use case.

Further Reading and Final Thoughts

Atlantis also recently commissioned a survey that was conducted by Scott D. Lowe at ActualTech Media. You can read the results of “From the Field: Software Defined Storage and Hyperconverged Infrastructure in 2016” here. It provides an interesting insight into what is happening out there in the big, bad world at the moment, and is definitely worth a read. Scott, along with David M. Davis and James Green, has also written a book – “Building a Modern Data Center – Principles and Strategies of Design”. You can reserve your copy here. While I’m linking to articles of interest, this white paper from DeepStorage.net on the Atlantis USX solution is worth a look (registration required).

I really like the focus by Atlantis on simplicity. Particularly if you’re looking to deploy these things in a fairly remote destination.

Secondly, the built-in resiliency of the solution allows for operational efficiencies (you don’t have to get someone straight out to the site in the event of a node failure). I also like the fact that you can use these as a starting point for a HCI deployment, without a significant up-front investment. Finally, the use of all-flash helps with power and cooling, which can be a real problem in remote sites that don’t have high quality data centre infrastructure options available.

I’ve been impressed with Atlantis in the discussions I’ve had with them, and I like the look of what they’ve done with the CX-4. It strikes me that they’ve thought about a number of different scenarios and use cases, and they’ve also thought about working with customers beyond the purchase of the first appliance. Given the street price of these things, it would be worthwhile investigating further if you’re in the market for a hyperconverged solution.