Atlantis ILIO – Diskless VDI & Services Feature Review
My first introduction to Atlantis ILIO was at a User Group meeting in the UK last year. I had a chat with Jim Moyle who explained the Atlantis ILIO product to me. Their primary focus is Virtual Desktop Infrastructure (VDI) storage and performance optimization solutions. In a nutshell, the ILIO software appliance sits in the I/O path between your hypervisor and storage and does what is essentially I/O acceleration. Since the appliance sits in the I/O path, it presents an iSCSI or NFS datastore at the front-end to the hypervisor, and at the back-end, it is presented with NFS, iSCSI or Fibre Channel storage from the storage array. The appliance does a number of things such as inline deduplication, and also does Windows I/O optimization through Atlantis’ patented technologies. They can identify the different types of windows I/O, and intelligently process and characterize it. This allows them to determine which I/Os are latency sensitive & which are not, and prioritize accordingly. Atlantis claims that this technology massively reduces virtual desktop I/O to the back-end storage. They also claim that their technology reduces storage consumption, and makes VDI cost-effective from a storage perspective without sacrificing desktop performance (which has always been the Achilles heel of VDI).
At VMworld 2012, I had the opportunity to once again catch up with Jim, and also with Anjan Srinivas, the Director, Product Marketing at Atlantis Computing. I suppose the first thing to mention is that Atlantis Computing won the best of VMworld 2012 award in desktop virtualization for their Atlantis ILIO Diskless VDI product. I met with Anjan to ask about this ‘Diskless VDI’ product that won the award.
Atlantis ILIO Diskless VDI
Anjan told me that a key innovation this year was to use RAM as primary storage for stateless virtual machine images. This now enables customers to build elastic VDI data centers with just servers and software, moving the VDI market towards software defined data centers. This not only improves the CAPEX equation but also significantly simplifies the OPEX and management in data center. In the past Atlantis ILIO worked in conjunction with either a SAN/NAS or a local drive, as shown here.
Through intelligently optimizing how Microsoft Windows operating systems interact with VDI storage, and Atlantis’ inline deduplication, the amount of space required by an individual desktop is now so efficient that RAM becomes the store for the non-persistent virtual desktop. So no back-end storage required. In fact, I can now add that Colt Technology Services won the “Best of VMworld 2012” Europe award for its Agilisys Desktop-as-a-Service (DaaS) project deployed with Atlantis ILIO Diskless VDI. That’s two awards at two VMworlds for Atlantis Computing and the Diskless VDI solution.
We then went on to discuss availability of the appliance, for instance when an ESXi host (or the appliance itself) has a crash. Since ILIO runs as an appliance on an ESXi host, it can use vSphere features like vSphere HA or FT, although according to Anjan, most customers use HA. For persistent desktops, there is no risk of data loss, which is the primary concern. ILIO, even though it sits on the I/O path, does not keep any data. There is a crash consistent file system with journaling in each appliance. For non-persistent environments, the desktops reside purely in RAM. VMware View will recognize that the desktops have gone down, and will makes those desktops available on another host. Users may experience a temporary disconnect, but will be reconnected when the desktop is re-instantiated with their profiles.
We then went on to talk about a few vSphere integration points. The first was around the management UI. ILIO Center is used to manage the appliance. It has a plugin to vCenter, and it appears as a separate tab.
Next we talked about VAAI integration. There isn’t any just yet, but Anjan pointed out that they are working towards VCAI (View Composer Array Integration) which is still in tech preview mode at the moment. Atlantis already has a technology called ILIO FastClone which is used to create full clone desktops (persistent or stateless) by manipulations of the metadata as opposed to working on actual data in just a few seconds. Anjan stated that this was something they began using internally for deploying large numbers of virtual desktops quickly for QA purposes, but very quickly realized that this is something that many customers would benefit from. Atlantis’ plan is to integrate this technology with VCAI so that when View Composer requests the creation of a desktop based on a linked clone, the clone creation overhead is handed off to Atlantis ILIO.
Atlantis ILIO Deployment Services
Anjan then gave me some breaking news from the Atlantis camp. He told me that they were on the verge of releasing a feature called Atlantis ILIO Deployment Services on October 24th. This was first announced at VMworld 2012. He stated that one of the primary concerns they have is that storage for VDI does not get deployed with optimal configuration & sizing, resulting in customers not seeing all the possible benefits from their solution. To address this, Atlantis ILIO Services provides fundamental integration with VMware View to automate the whole VDI deployment process in a few clicks. There are basically 4 steps involved:
1. Auto deploy ILIO on any number of ESXi servers
2. Sizes & configures the environment accordingly, selecting the optimal density of virtual desktops
3. Creates and registers a datastore with vSphere
4. Hands the datastore off to View Composer for the deployment of virtual desktops
The bottom line is that this automation of deployment should remove any user errors from the setup, and give customers the optimal ILIO deployment from the start as per Atlantis Computing best practices.
This is very impressive technology, and it is all done in software. Atlantis are really opening the doors for many customers who have not been able to pursue VDI due to storage costs either due to capacity or performance. And they certainly deserve those VMworld accolades.
Get notification of these blogs postings and more VMware Storage information by following me on Twitter: @CormacJHogan
8 Replies to “Atlantis ILIO – Diskless VDI & Services Feature Review”
I am still wondering from a data consistency perspective how this is all handled. Especially when a host fails and data is “in RAM” and not committed to disk yet. I can’t find any details around this surprisingly enough.
If I am not mistaken it is an “in RAM deduped coalesced write back” solution. So what happens if stuff breaks.
I work for Atlantis and I’d be happy to answer your question. The “in RAM deduped coalesced write back” is mainly deployed for persistent desktops and in this case we have a transactional file system turned on to ensure that the vms com back up in a consistent state.
The Atlantis Diskless for VDI solution where everything is done in RAM and nothing is written back, is for stateless desktops and in this case the user just logs back on again to a different desktop and their apps and persona are recreated. Nothing on ILIO needs to be kept.
Let me know if this isn’t clear and I’d be happy to have a much more detailed conversation.
Thanks for the info Jim, much appreciated.
yeah but if you take another transactionnal filesystem with inline deduplication like zfs, to maintain a good performance ration and consistancy of data they use a special stack called ZIL or SLOG on ssd to maintain the consistency state.
You can have a transactionnal filesystem with journaling write back and everything you want if you use ram you will always loose data
Great post Cormac. Just wanted to clarify one thing around the following quote: “VMware View will recognize that the desktops have gone down, and will makes those desktops available on another host.”
View will recognize that the desktop’s are unavailable, but it will not deploy extra desktops until the “missing” desktops are removed from the View ADAM database.
I believe Atlantis have a set of scripts that you can run when there is a host failure that queries vSphere to find out which desktop vm’s have been lost. It will then proceed to remove the missing desktops from ADAM. Once that is complete, View will then deploy new desktops replace those that are lost.
Ah – very interesting. Thanks very much for the clarification Simon.
Comments are closed.