VMware changed the economics of the x86 server market when it introduced its ESX hypervisor in 2003. Through virtualisation users got closer control over their resources, improved server utilisation and spent a lot less on server hardware as a result. Initially server hypervising was used for consolidation, helping to cut down the number of physical servers and data centre locations. In more recent times the separation of physical from virtual resources has inspired new usage models and driven the move towards Cloud Computing.
Successful server hypervisors have created their own ecosystems: VMware, Citrix (Xen), Microsoft (Hyper-V) and RedHat (KVM) have thousands of attached partners who address their APIs and build plugins into their systems management suites. Each environment also creates a grouped set of aggregated resources, with various players acting together – witness IBM’s set of services with VMware for instance. Dell, HP and IBM jockey for position as VMware’s largest partner, underlining the power of its business model.
So if server hypervisors are commercially important, what about storage hypervisors? Despite being at a very early stage of development, there are a number of important issues. In particular:
- Most are themselves currently subservient to specific server hypervisors – Virsto for instance attaches specific offerings to Hyper-V (server and VDI separately) and vCenter/ESX (together)
- They offer a new way for CIOs to reduce the extra costs of storage associated with virtualisation; over time they may genuinely change the economics of storage
- They are being positioned against typical storage systems vendors who offer savings only for customers of their own offerings – analogous to server vendors before 2003: Hitachi and IBM are exceptions in allowing other vendor arrays to be connected to their storage hypervisors; in the server market the ability to run hypervisors on any vendor’s servers was a key to acceptance
- There is a growing debate about whether a software-only approaches from Falconstor, Data Core and Virsto are better than hardware and software ones from IBM and Hitachi – similarly their were a number of software and hardware approaches to x86 server virtualisation before VMware’s software-only approach succeeded
Of these issues the first is important – after all if each storage hypervisor is relevant only to specific server virtualisation suites, then perhaps VMware itself will turn out to be the ‘VMware of storage hypervisors’. A generic market will depend on storage virtualisation being handled as a separated management layer for pools of storage as envisioned in Utility Computing models of the late 1999s. It is clear that server hypervisors have already achieved that role for server pooling.
Worldwide spending on storage systems stands at $120 billion today and all vendors offer savings through new techniques, not least by the development of automatic tiering and the attachment of low-cost disk in Tier 3: however the development of successful storage hypervisors looks to us to have the best chances of creating a step change in storage economics. Watch out for when any of the storage hypervisor players start recruiting partners en masse – it will be a good sign of success for the subject as a commercial entity.
Read more on how we assess go-to-market rainmakers.