Michael Theroux <mtheroux2 <at> yahoo.com> writes: > > Hello, > A number of weeks ago, Amazon announced the availability of EBS Optimized instances and Provisioned IOPs for Amazon EC2. Historically, I've read EBS is not recommended for Cassandra due to the network contention that can quickly result (http://www.datastax.com/docs/1.0/cluster_architecture/cluster_planning). > > Costs put aside, and assuming everything promoted by Amazon is accurate, with the existence of provisioned IOPs, is EBS now a better option than before? Taking the points against EBS mentioned in the link above: > > > EBS volumes contend directly for network throughput with standard packets. This means that EBS throughput is likely to fail if you saturate a network link. > According to Amazon, Provisioned IOPS guarantees to be within 10% of of the provisioned performance 99.9%of the time. This would mean that throughput should no longer fail. > EBS volumes have unreliable performance. I/O performance can be exceptionally slow, causing the system to backload reads and writes until the entire cluster becomes unresponsive. > Same point as above. > Adding capacity by increasing the number of EBS volumes per host does not scale. You can easily surpass the ability of the system to keep effective buffer caches and concurrently serve requests for all of the data it is responsible for managing. > I believe this may be still true, although I'm not entirely sure why this is more true for EBS volumes vs. emphemeral. > > Any real world experience out there with these new EBS options? > > -Mike > > >
Thanks for asking this as I'm wondering the same thing. I'd like to hear other people's take on this as I'm trying to spec out a Cassandra ring on AWS.
