Atlantis Computing have added a new, smaller form-factor hyper-converged product to their existing CX line, the CX-4.
Designed with Remote Office/Branch Office (ROBO) deployments in mind, the CX-4 is a two-node design, specifically created to keep the price down on deployments that can’t justify the more expensive CX-12 or CX-24 units.
The CX-4 is an all-flash unit, whereas other hyper-converged aimed at ROBO deployments tend to be spinning disk or hybrid setups. According to Atlantis, they looked at a hybrid of disk and flash, but the cost difference was only about 5% for equivalent usable capacity, so they just went with all flash.
“The new Atlantis HyperScale CX-4 truly meets an unmet need in the industry and is a game changer for our ROBO customers,” said Chetan Venkatesh, president and CEO of Atlantis. “Prior to this offering, it would be unheard of for a ROBO environment to be equipped with an all-flash hyper-converged appliance because of the cost.”
The CX-4 is available on physical hardware from SuperMicro, [entity display=”Dell” type=”organization” subtype=”company” active=”true” activated=”true” deactivated=”false” key=”dell” natural_id=”fred/company/1260″]Dell[/entity], HP Enterprise, [entity display=”Lenovo” type=”organization” subtype=”company” active=”true” activated=”true” deactivated=”false” key=”lenovo” natural_id=”fred/company/91366″]Lenovo[/entity], and Cisco, but of particular interest to me is the use of the [entity display=”Dell” type=”organization” subtype=”company” active=”false” key=”dell” natural_id=”fred/company/1260″]Dell[/entity] PowerEdge FX2 server platform option. The FX2 has an inbuilt redundant networking capability, meaning you don’t need external networking devices to connect the nodes together. The FX2 is a more expensive option than the SuperMicro TwinPro unit that Atlantis are using for their low cost option of about US$43,000 list, but the saving on additional switch infrastructure is important to factor in.
Usually with hyper-converged devices, you need some form of external networking connection to connect nodes with one another so they can perform their scale-out functions. The CX-4 has the ability to directly cable the two-node option together, but if you need to add nodes, you’ll need to go buy a couple of switches so they can all talk to each other. The switches can’t be some super-low-end devices either, because of the performance requirements to keep the nodes synchronized, so the total cost of a deployment needs to factor them in.
External switches also add complexity to the deployment, and when you need to scale to many sites — such as in a retail chain, or remote oil and gas rigs — adding complexity to sites that often don’t have onsite support staff is a huge management challenge. Simple and reliable is the name of the game, which is why I like the idea of having as few devices and cables as possible.
Hyper-converged is shaping up to be a major battleground for vendors, so expect to see a lot of product announcements like this in coming months.
This article first appeared in Forbes.com here.