Atlantis Computing: A New Approach

When it comes to customers considering Virtual Desktop Infrastructure (VDI) deployments, many will end up “bolting on” a VDI workload to existing storage that is servicing mixed workloads such as databases, file services etc. This is likely to be mid-range VNXe, FAS or other.

When VDI is “added” to that mix it can result in one possible outcome…mayhem. That’s primarily from a performance and manageability perspective.

Enter Atlantis ILIO, with a new approach which is about ripping up the instruction book and starting again.

Atlantis Value Proposition

Atlantis started with a strategy to target VDI, and specifically two types of solutions targeting the Enterprise:

Persistent Desktops
Stateless deployments

The approach is to use server RAM to accelerate IO and specifically to “de-risk cache sizing”. This should result in at worst-case 5x VMs running on the same storage, and possibly as much as 8-10x.

Within the Atlantis customer base they claim to have seen:

74% virtualization of customer base.
40% increase in storage capacity every year…. doubling storage every two year.

We’re not talking about a “Flash-first” strategy. This is using server-side memory, which has been getting cheaper and cheaper, as the primary storage tier. This is kind of similar to vendors who have bet on Moore’s law for CPU. Data Domain comes to mind. So as time progresses and we move closer to Server Class Memory (SCM), using server RAM will become the primary storage tier.

One thing to note is that products like VSAN, Pure, COHO Data who we have met so far on VFD3 all use some type of Flash/SSD as the Tier-0. The mechanics of how and what depend on each vendors architecture.

Deployment Models

For Atlantis this is how they do it:

Everything can be run in memory and this doesn’t have to be backed by any storage. This is one use case and is the “stateless” scenario.
- However you should know that you can back this off to disk. In this scenario you let Atlantis take care of creating a large virtual “filesystem”. They provide the abstraction and intelligence.
- This is how you balance performance and capacity. Memory provides very fast IO and spinning disk provides capacity. You could use Flash as the second tier either for super-fast solution.
- You create policy driven configurations based on your requirements and Atlantis does the rest.

Atlantis is about scale, with large provisioning engines at work. This is how Atlantis has a customer like JP Morgan Chase with 100,000 desktops across 19 data centers.
Atlantis claim that you can achieve better performance for storage, with reduced risk.

In terms of platform support, Atlantis is supported across vSphere, Hyper-V and XenServer. In terms of backing off the RAM to actual disk, you can do this and it’s supported by EMC, IBM, Netapp, HDS, SSDs etc.

Evolution into server workloads

Atlantis tackled VDI first and have proven their capability in this space. They have 500,000 VMs across their over 400+ customers.

They can also talk about 100% YOY growth, with 261 customers added in the last year. One important point to note is that 92% of their revenue is achieved via a channel strategy and via “Global Partners” such as EMC, HP, IBM etc..

They have just released a product designed specifically for virtualised server workloads …. based on their experience with use of memory to meet VDI challenges. This is called USX and is designed to decouple applications from storage, with capacity presented via a “persistent tier” and performance from in-memory storage.

Chew that data first

Now for the optimising piece …..

We also also talking about the following additional features applied in-memory:

In-line deduplication
Compression
I/O Coalescing (occurs before writing to disk)
Thin Provisioning (occurs in memory)

So the data is optimised before hitting the spinning disk. Like any deduced solution, this allows economies of reduced scale be applied to the smaller post-process dataset. So this means less data on the network which is always what it should be about.

Interoperability

One thing that impressed me was they want to embrace VSAN. Not that they shouldn’t offer a competing product but there is some bashing going on from different vendors of other vendors products. That’s just an example. They are also about to launch an Openstack Swift and Cinder plugin in the next 2-3 months. So from an interoperability perspective Atlantis are making the right noises. My view is that this is what customers want to hear !. Whether they are actually better or not, or whether VSAN is required is irrelevant – they are providing investment protection for their customers.

It’s clear that anyone considering VDI should consider Atlantis as a vendor we will hear a lot about over the coming years. I suggest you check them out further at their website here www.atlantiscomputing.com