Speedometer 2014Yesterday I wrote a blog post concerning the availability of the beta version of Unitrends release 7.3 for our Recovery-series physical backup appliances and our virtual appliances (Unitrends Enterprise Backup™.)  Today I thought I’d talk a little bit about various performance improvements that have been made in release 7.3.  As a reminder, release 7.3 beta is available now – the general release of 7.3 will be in late December.

What’s interesting about performance is that it’s a never-ending journey.  You’ve got to increase performance month-by-month, quarter-by-quarter, and year-by-year – because we’re in an industry where storage is always increasing and where market forces mean that we have to protect increased storage for less.  One way we do that is releasing new generations of backup appliances – which we did in October – and another way we do that is constantly optimizing our software.

With that said, here’s a list of some of the major performance improvements in release 7.3:

Faster  Windows 2012 R2 Hyper-V backups residing on CSV (Clustered Shared Volumes).   Windows 2008 versions of Hyper-V force the sequential backup of Hyper-V virtual machines due to  limitations in the Microsoft VSS (Volume Shadow Copy Service) data protection framework.  With Hyper-V 2012 R2,  the CSV implementation is now updated to support multiple VMs being backed up at the same time.  The ability to backup multiple VMs at once will significantly improve the backup window associated  with Windows 2012 R2 Hyper-V VMs. This improves backup performance in many cases by 200% or more.

Improved deduplication rates and replication performance for VMware vSphere.  In VMware vSphere environments in which there is a high change rate, there have been optimizations made to our deduplication capabilities that result in higher retention and faster replication due to deduplication improvements.  Replication performance in these environments can increase by 50% or more.

Deduplication signature  algorithm improvements.  A deduplication signature is a representation of a segment of data – it’s computed by reading through all of the data and computing a unique identifier that can then be used to quickly figure out if that block of data already exists on the system.  In release 7.3, we’re using a new technique that typically is double the performance of prior techniques.  Now – signatures are only one part of deduplication – so that translates into 10% faster deduplication and replication.

Faster replication of synthetic backups.  Synthetic backups are the result of periodic on-appliance “masters” when incremental forever is chosen as the backup strategy to protect physical or virtual systems.  We’ve optimized the communication with respect to synthetics and their associated groups between the replication source (the local backup appliance) and the replication target (the disaster recovery appliance) such that synthetics replication often occurs about 40% faster.

Faster VMware vSphere restore screen user experience.  We’ve rewritten and optimized the scan of the VMware vSphere environment to query only the required elements.  This practically means an improved scan time from 200% to 500% and could be as much as an order of magnitude improvement in large VMware vSphere environments with a lot of virtual machines (VMs.)

There’s more – but that’s enough for now.  Hope everyone has a Happy Thanksgiving!  And as always, let us know what you think.

[Special thanks to Sameer Kamat, our architectural leader, for helping with a lot of the details here.  And of course thanks to our entire engineering team, led by Joe Balazs, for all of the work to provide both the features discussed in the prior post and the performance discussed in this post.]