Scale-out with MySQL Cluster
Update: webinar replay is now available from http://event.on24.com/r.htm?e=311660&s=1&k=3DCFE1CB3E1CF3F0FD0969DC66D93989
On Thursday 26th May Mat Keep and I will be presenting a webinar on how MySQL Cluster can deliver linear scalability – together with some tips on how to achieve it. As always the webinar is free but you need to register here.
The session starts on Thu, May 26 at 17:00 UK time, 18:00 Central European Time, 09:00 Pacific.
This webinar will discuss best practices in scaling services on-demand for high volumes of reads and writes, and provide insight on the range of NoSQL and SQL access methods available to developers, specifically covering:
- Automatic partitioning (sharding) for high scalability
- On-line scaling of the cluster across commodity hardware
- SQL and NoSQL interfaces, and what should be used when
- On-line updating of schema design to accommodate rapidly evolving applications
- Resources to get started
Looks like your graph actually shows the impossible: better than linear scaling. Updates more than double from 8 to 16 nodes. Can you post the actual numbers used to generate this graph?
Hi Baron,
the oddity actually lies with the sub-linear growth between 4 and 8 nodes – smooth that out and you get nice linear write-scaling from 4 to 16 nodes. Unfortunately we only had a very short time in this particular benchmark environment and so we haven’t been able to understand what happened in the 8 data node run or retry it. We do plan on getting back into this benchmark lab and of course I’ll share the results.
Andrew.
Actual numbers are here: 1.15M, 2.13M and 4.33M reads and 682k, 987k and 2.46M updates for 2, 4 and 8 nodes respectively.