In the year of buzzwords, marketing FUD (fear, uncertainty and doubt) and dramatic changes in industry architecture, it can be hard to see the wood for the trees.
The most common questions I received in 2016 all surrounded the use of SAN and hyperconverged technologies. Questions such as "Should I be going hyperconverged?" and "Do I still need a SAN?" came up time and time again. I wanted to use this post to try and clear up some of the reasons for, and against SAN technology, as well as delve into how the latest advances in hyperconverged technology can support strategic business decisions.
For anyone who hasn't assessed a storage platform in the past couple of years, hyperconverged platforms aggregate local storage from multiple servers into a RAIN (Redundant Array of Independent Nodes) architecture. This is very similar to RAID, however, the mirroring of data between disks is happening between the nodes. This turns all storage into a single pool, which can be accessed from multiple servers, whilst also retaining resilience. Whilst different vendors have a variety of proprietary methods for providing this service, the nuances of each method is perhaps a topic for another day.
Until very recently, the arguments for and against Hyperconvergence were actually very simple and came down to a few key points:
These questions really focus us in on what infrastructure design choices are available to your organisation. Further to this, in the current marketplace, on-premise infrastructure design is mostly a choice between traditional SAN and host infrastructure, or hyperconverged infrastructure.
The two latest software releases from VMware and Nutanix are major game changers in the assessment of hyperconverged technologies. The technologies in question are Nutanix Acropolis Block Services and the VMware Virtual SAN iSCSI Target Service. These technologies can completely remove the requirement for a SAN by allowing the pool of storage within a hyperconverged platform to be utilised for physical and clustered workloads.
Prior to these advancements, the requirements for LUN's or shared storage would force infrastructure to retain some form of shared storage. In the case of Oracle and SQL licencing, the entire virtual cluster must be licensed, causing you to incur a major expense if you only require a couple of SQL servers. Moving forward an option could be to take a single virtual host, or a couple of physical servers, and utilise the storage pool available within a hyperconverged platform, thus gaining higher performing storage.
For example, take a heavily virtualised infrastructure running on a combination of hosts and storage with a balanced compute and storage requirement.Assuming our example organisation has around four virtual hosts running hypervisor technology, these hosts receive all of their storage from a SAN solution. The organisation also has one or two physical servers which require block based storage from a SAN technology.
In this case, hyperconverged technology is a perfect fit and could be deployed using a single 4 node 2U hyperconverged appliance utilising block storage technology to supply any servers outside of the virtual environment. This is a paradigm shift in the way hyperconverged use cases work, previously they were only fit for general purpose virtualisation and VDI, meaning the true benefits could not be realised for many organisations.
This deployment consolidates both the SAN and servers into a single 2U appliance the potential ROI could be enormous:
Now this doesn't mean that hyperconverged infrastructure is the perfect fit for every use case, there are scenarios where extreme IO performance is required, or where the infrastructure is so storage-heavy that the infrastructure blocks don't scale evenly enough to make a healthy use case.
Nevertheless, to put it simply, these new technologies blur the lines between a hyperconverged platform and a SAN technology, distorting engineered infrastructure systems and hyperconverged infrastructure appliances. The SAN may be on its way out for many use cases, however, shared storage is not going away anytime soon, not when many people are rushing out to deploy an iSCSI based block for running a shared storage solution.
Share this page: