Before we get too far into this post let’s first get some terminology straight. StorMagic refers to your remote or branch offices as the “edge” – This might help when reading through a lot of their marketing material as sometimes I tend to relate “edge” to networking, more specifically entry/exit points.
StorMagic, a UK based company was founded in 2006 and set forth to develop a software based storage appliance that enterprise can use to solve one big issue – shared storage at the edge. StorMagic is another one of those companies that presented at VFD4 that really had a strong sense for who their target markets are – they aren’t looking to go into the data center (although there is no reason that they can’t), they aren’t looking to become the end all be all of enterprise storage (although I’m sure they would love that) – they simply provide a shared, highly available storage solution for those companies that tend to have many remote branch offices with a couple of ESXi (or Hyper-V) hosts. On the second day of VFD4 in a conference room at SolarWinds, StorMagic stood up and explained how their product, SvSAN can solve these issues.
Another VSA but not just another VSA
Choosing between deploying a traditional SAN vs. a VSA is a pretty easy thing to do – most the times it comes down to the shear fact that you simply don’t have enough room at your remote site to deploy a complete rack of infrastructure, nor do you have the resources on site to manage the complexity of a SAN – so a VSA presents itself as a perfect fit. With that said there are a ton of VSA’s out on the market today so what is setting StorMagic apart from all the other players in the space? Why would I chose SvSAN over any other solution? To answer these questions let’s put ourselves into the shoes of a customer in StorMagic’s target market – a distributed enterprise with anywhere from 10 to 10000 remote “edge” offices.
One of the driving forces behind SvSAN’s marketing material is the fact that you can setup your active/active shared storage solution with as little as 2 nodes. 2 – Two. Most VSA vendors require at least a three node deployment, and justifiably they do this to prevent a scenario called split-brain. Split-brain is a scenario where nodes within a clustered environment become partitioned, with each surviving node thinking it’s active, which results in a not so appealing situation. So how does StorMagic prevent split-brain scenarios with only two nodes? The answer lies in a heartbeat mechanism called the Neutral Storage Host (NSH). The NSH is recommended and designed to run centrally, with one NSH supporting multiple SvSAN clusters. Think one NSH supporting 100 remote SvSAN sites. The NSH communicates back and forth with the SvSAN nodes in order to determine who is up and who is down, thus being the “tie breaker” if you will in the event the nodes become partitioned. That said the NSH is an important piece to the SvSAN puzzle and it doesn’t necessarily need to run centralized. For those sites that don’t have good or any bandwidth, the NSH can be run on any Windows, Linux, Raspberry Pi device you want, locally at the site. Beyond the heartbeat mechanisms of the NSH, SvSAN also does a multitude of things locally between the two nodes to prevent split brain as well. It can utilize any one of its networks, be it management, iSCSI, or mirroring network to determine and prevent nodes from becoming partitioned. So with all this what advantages come from not requiring that third node of compute within the cluster – well, one less VMware license, one less piece of hardware you have to buy and one less piece of infrastructure you need to monitor, troubleshoot and backup – which can add up to a pretty hefty weight in loonies if you have 10000 remote sites.
Aside from lowering our infrastructure requirements SvSAN brings a lot of enterprise functionality to your remote sites. It acts in an active/active fashion, synchronously replicating writes between each node. When a second node of SvSAN is introduced, a second path to the VMs storage is presented to our hosts. If at any time one host fails, the other host containing the mirrored data can pick up where it left off, which essentially allows VMware HA to take our VMs that were running on local storage on the failed host, and restart them on the surviving host using local storage. While the failed node is gone, the surviving SvSAN journals and writes meta data around the changes that occur in the environment, minimizing the time that it will take to re-synchronize when the original node returns. That said the original node isn’t required for re-synchronization – the benefits of the SvSAN architecture allow for the second node to come up on different hardware or even different storage. This newly added node will be automatically configured, setup and re-synchronized into the cluster, same goes for the third, the fourth, the fifth node and so on, with just a few clicks.
As far as storage goes, SvSAN can take whatever local or network storage you have presented to the host and use that as their datastore. The appliance itself sits on a datastore local to the host, somewhere in the terms of 100GB – from there, the remaining storage can be passed straight up to SvSAN in a JBOD, RDM, or “vmdk on a datastore” fashion. SvSAN also gives us the ability to create different storage tiers, presenting different datastores to your hosts depending on the type of disk presented, be it SATA, SAS, etc. In terms of SSD, SvSAN supports either running your VMs directly on solid state datastores, or you can carve up SSD tier to be used as a write-back cache to help accelerate some of those slower tiers of storage.
In terms of management, StorMagic is fully integrated into the vSphere Web Client via a plug-in. From what I’ve seen, all of the tasks and configuration that you need to perform are done through very slick, wizard driven menus within the plug-in, and for the most part StorMagic has automated a lot of the configuration for you. When adding new nodes into the VSA cluster, vSwitches, network configurations, iSCSI multipathing – they are all setup and applied for you – when recovering existing nodes, surviving VSA’s can push configuration and IQN identifiers down to the new nodes, making the process of coming out of a degraded state that much faster.
Wait speaking of VMware
Worse transition ever but hey, who better to validate your solution than one of the hypervisors that you run on. As of Feb 4th, VMware and StorMagic have announced a partnership which basically allows customers to couple the new vSphere ROBO licensing with a license for SvSAN as well. Having VMware, who took a shot at their own VSA in the past (ugh, remember that!) chose your product as one they bundle their ROBO solutions with has to be a big push of confidence for both StorMagic and their potential customers. You can read more about the partnership and offering here – having both products bundled together is a great move on StorMagic’s part IMO as it can really help push both adoption and recognition within the VSA market.
Should I spend my loonies on this?
IMO StorMagic has a great product in SvSAN – They have done a great job in stating who their target market is and who they sell to – and defending questions to no end with that market in mind. HA and continuous up time is very important to those enterprises that have distributed architecture. They’ve placed these workloads at the “edge” of their business for a reason, they need the low latency, and honestly, the “edge” is where a company makes their money so why not protect it. With that said I see no reason why an SMB or mid market business wouldn’t use this within their primary data center and/or broom closet and I feel they could really benefit by maybe focusing some of their efforts in that space – but that’s just my take, and the newly coupled VMware partnership, combining SvSAN with the ROBO licenses kind of de-validates my thinking and validates that of StorMagic – so what do I know . Either way I highly recommend checking out StorMagic and SvSAN for yourself – you can get a 60 day trial on their site and you can find the full library of their VFD4 videos here.