Comparing VMware Virtual SAN with EMC ScaleIO and conventional storage arrays

Software-defined and hyper-converged storage solutions are now a viable alternative to conventional storage arrays so let’s take a quick look at how two of the most popular solutions compare – VMware Virtual SAN (VSAN) and EMC ScaleIO:

Architecture

On vSphere this is an easy win for VMware as VSAN is delivered using kernel modules which provides the shortest path for the IO, has per Virtual Machine policy based management and is tightly integrated with vCenter and Horizon View.

ScaleIO is delivered as Virtual Machines, which is not likely to be as efficient, and is managed separately from the hypervisor – on all other platforms ScaleIO is delivered as lightweight software components not Virtual Machines.

VSAN also has the advantage of being built by the hypervisor vendor, but of course the downside of this is that it is tied to vSphere.

Availability

Win for EMC, since the failure of a single SSD with VSAN disables an entire Disk Group. Although VSAN has the ability to support up to three disks failures where as ScaleIO only one, in reality the capacity and performance overhead of supporting more than one failure means that VSAN will nearly always be used with just RAID 1 mirroring.

If you need double disk failure protection you are almost certainly better off using a storage array.

Performance

Easy win for VMware as VSAN uses SSDs as a write buffer and read cache, ScaleIO does have the ability to utilise a RAM read cache.

Flexibility

Easy win for EMC as with ScaleIO you can:

  1. Utilise physical servers running Windows and Linux
  2. Utilise hypervisors running vSphere, Hyper-V, XenServer and KVM
  3. Utilise any storage supported by the OS or hypervisor
  4. Utilise any combination of HDDs and SSDs as required
  5. Create multiple Protection Domains per system for greater resiliency
  6. Create Storage Pools for each storage tier within a Protection Domain
  7. Mix and match nodes with dissimilar configurations

VSAN has a more rigid architecture of using Disk Groups which consist of one SSD and up to seven HDDs.

Elasticity

Easy win for EMC as ScaleIO supports up to 1,024 nodes, 256 Protection Domains and 1,024 Storage Pools, and auto-rebalances the data when storage is added or removed.

ScaleIO can also throttle the rebuilding and rebalancing process so that it minimises the impact to the applications.

Advanced Services

Easy win for EMC as ScaleIO provides Redirect-on-Write writeable snapshots, QoS (Bandwidth/IOPS limiter), Volume masking and lightweight encryption.

Licensing

This is a tricky one as VSAN has the more customer friendly licensing as it is per CPU therefore as new CPUs, SSDs and HDDs are released you will be able to support more performance and capacity per license.

ScaleIO has a capacity based license which is likely to mean that further licenses are required as your capacity inevitably increases over time. There is also two ScaleIO licences – Basic and Enterprise (adds QoS, Volume masking, Snapshots, RAM caching, Fault Sets and Thin provisioning).

The one downside of VSAN licensing is that you need to licence all the hosts in the cluster even if they are not used to provision or consume VSAN storage.

Conventional storage arrays

What are the advantages of a conventional mid-range array?

  1. Rich data services – most storage arrays include de-duplication, compression and tiering along with many other advanced features
  2. Unified storage – many storage arrays support both block and NAS protocols
  3. Replication – many storage arrays support synchronous and metrocluster solutions
  4. Integrated data protection – some storage arrays do not require a separate backup solution
  5. Usable capacity – most storage arrays support parity RAID which can achieve usable capacity ratios of up to 80%
  6. Double disk protection – whilst this is supported on VSAN it is almost certainly not practical at scale
  7. Turnkey solution – with a single contact for support of all hardware and software

What are the advantages of hyper-converged software-defined solutions?

  1. Multi node failure – can tolerate the failure of more than one node
  2. Rapid rebuilds – as they take place in parallel across multiple drives
  3. Bring your own hardware – take advantage of commodity prices
  4. Built-in “IT Deflation” – as over time hardware unit costs drop
  5. Independent – the software lives on beyond the life of the hardware
  6. Elasticity – non-disruptively grow and shrink as required
  7. Low ongoing costs – perpetual license followed by annual maintenance
  8. Gain new features – just by upgrading the software
  9. Simplified management – compute and storage managed together

So which is best?

As always each vendor will build a strong case that their solution is the best, in reality each solution has strengths and weaknesses, and it really depends on your requirements, budget and preferences as to which is right for you.

For me the storage array is not going away, but it is under pressure from software-defined and cloud based solutions, therefore it will need to deliver more innovation and value moving forward. The choice between VSAN and ScaleIO really comes down to your commitment to vSphere – if there is little chance that your organisation will be moving away, then VSAN has to be the way to go, otherwise the cross-platform capabilities of ScaleIO are very compelling.

As always feedback would be welcome.

Related Posts

  1. An introduction to VMware Virtual SAN
  2. VMware EVO:RAIL or VSAN – which makes the most sense?
  3. An introduction to EMC ScaleIO
  4. What are the pros and cons of Software-Defined Storage?
  5. Lock-in, choice, competition, innovation, commoditisation and the Software-Defined Data Centre
Mark Burgess has worked in IT since 1984, starting as a programmer on DEC VAX systems, then moving into PC software development using Clipper and FoxPro. From here he moved into network administration using Novell NetWare, which kicked-off his interest in storage. In 1999 he co-founded SNS, a consultancy firm initially focused on Novell technologies, but overtime Virtualisation and Storage. Mark writes a popular blog and is a frequent contributor to Twitter and other popular Virtualisation and Storage blog sites.
twittergoogle_pluslinkedinmailtwittergoogle_pluslinkedinmailby feather

About Mark Burgess

Mark Burgess has worked in IT since 1984, starting as a programmer on DEC VAX systems, then moving into PC software development using Clipper and FoxPro. From here he moved into network administration using Novell NetWare, which kicked-off his interest in storage. In 1999 he co-founded SNS, a consultancy firm initially focused on Novell technologies, but overtime Virtualisation and Storage. Mark writes a popular blog and is a frequent contributor to Twitter and other popular Virtualisation and Storage blog sites.

Leave a Reply