If you have been following the virtualization blogosphere
then you may have noticed that last week we published three interesting
performance whitepapers at once. I want to talk about the “VMware vSphere 4
Fault Tolerance: Architecture and Performance” whitepaper in this blog.
VMware Fault
Tolerance also known as FT is one of the most anticipated features of the
vSphere 4.0 release. FT provides continuous high availability to virtual
machines without down time or disruption in the event of complete host failure
and it is designed to be agnostic to the guest operating system and the
application running in it. With FT you could get hardware style fault tolerance
protection to virtual machines on commodity server systems without the need for
any specialized hardware. All you need is recent processors from Intel and AMD
and a gigabit Ethernet Link for transmitting FT logging traffic. You could
download and use the VMware Site Survey
utility or check KB article
1008027 to see which servers in your datacenter have FT compatible
processors.
We first demonstrated a prototype version of FT at the VMworld 2007
keynote, and earlier this year at VMworld Europe 2009 we previewed its performance
characteristics. The product is now officially
shipped with vSphere 4.0 and some of you
may be already using the feature and many others are probably planning to use
it in their datacenters very soon. In case if you are wondering how enabling FT
impacts performance we got you covered in the whitepaper. This whitepaper
provides concise but important architectural and performance aspects of
enabling FT backed up with performance data from variety of workloads. We have
been working very hard on enhancing the performance of FT for the last few
years and you’ll be able to see the results in the whitepaper. The two key takeaways that you will learn
from this paper are that you don’t need lot of networking bandwidth for FT; a
gigabit link is sufficient for the vast majority of workloads. Also when there is sufficient CPU headroom FT
impacts throughput very little. For more
details I recommend you to take a look at the whitepaper.
I also welcome you to my VMworld 2009 session in San Francisco
titled “BC2961:
VMware Fault Tolerance Performance and Architecture” where I will be
talking more in depth with recent performance numbers.