Taking VMware vSphere Storage APIs for Data Protection to the Limit: Pushing the Backup Performance Envelope
VMworld 2012 in San Francisco has come and gone. VMworld EMEA in Barcelona is next week. It is a great event for technical audience; filled with numerous educational sessions, the largest solution pavilion and a lot of super awesome parties! VMworld issues call for papers each year. These are opportunities for you to secure a session to show off your innovative, unconventional or best practice methods for integrating VMware technologies. If your paper is approved, you get a speaker pass to the show for free. Remember these are not regular breakout sessions with product pitches you get as part of being a sponsor for the event; you get these because you have something valuable to share with technical savvy audience.
George Winter (Staff Technical Product Manger in Symantec, a respected virtualization evangelist), Roger Andersson (Sr. Technical Marketing Director in Cisco, an expert in virtualization hardware) and I brainstormed last spring as to what we could do for VMworld 2012. You may remember a joint performance benchmark from Symantec and Cisco in 2010 where we focused on vStorage APIs for Data Protection (VADP) mainly from SAN transport perspective. Two years later, SAN transport still remains the most popular VMware vSphere backup method for enterprise, as it is the only true offhost technology. With the increased popularity for NFS datastores and 10 Gigabit Ethernet, we thought we should do a performance benchmark on VADP again with focus on Network Block Device (NBD) and hotadd methods. The goal is to push VADP to the limit in such a way that it has the least impact on production virtual machines. On our way, we will also discover best practices to get the most out of VADP based backups.
We submitted the idea to VMworld with our plan. Not only did our submission got approved, VMware wanted to partner with us when the final white paper on this topic comes out! George and I decided to use NetBackup 5220 as the VADP backup device since it has everything needed for VADP built-in, Roger offered to bring the best of breed Cisco UCS blades for vSphere and Nexus for network gear. How about the production storage? Since NFS datastore is where the need for NBD and hotadd plays major roles, NetApp’s storage fits the bill. We reached out to our friend Mike Spinder (Practice Manger in Datalink; a joint partner for Cisco, NetApp, Symantec and VMware) for NetApp storage and hardware. He offered us to use the Datalink lab in Eden Prairie, Minnesota. Plus, we had the support from numerous NetApp storage experts from Datalink.
Well, the rest is history. A project that was meant as a benchmark and best practice development turned out to be an eye opener for the way NBD works and how to effectively utilize 10 Gigabit Ethernet for backups. We shared and validated results with VMware engineering team. We presented the details in VMworld 2012 in San Francisco.
The presentation is available for you on demand in virtual theatre from VMworld website. If you have VMworld 2012 subscription, we invite you to view the session here:
Roger Andersson, Cisco
Abdul Rasheed, Symantec
George Winter, Symantec
Use the View This Session button in the middle of the page.
Don't have VMworld 2012 subscription? A lite version of this session was done through this free webcast:
If you have questions or comments, feel free to add to this blog as comments. I shall respond to them as early as I can.