Perspectives - Technologies for Data Intensive Computing

Wednesday, November 11, 2009

Technologies for Data Intensive Computing

In an earlier post Andy Bechtolsheim at HPTS 2009 I put my notes up on Andy Bechtolsheim's excellent talk at HPTS 2009. His slides from that talk are now available: Technologies for Data Intensive Computing. Strongly recommended.

James Hamilton

e: jrh@mvdirona.com

w: http://www.mvdirona.com

b: http://blog.mvdirona.com / http://perspectives.mvdirona.com

Wednesday, November 11, 2009 4:45:37 PM (Pacific Standard Time, UTC-08:00)

Comments [2] - Trackback
Hardware

Wednesday, November 11, 2009 8:53:11 PM (Pacific Standard Time, UTC-08:00)

Do you think that development of SSD will extend lifespan of conventional SQL DBs and render "Big Data/nosql" storage solutions obsolete?
By "Big Data/nosql" I mean things like Google BigTable, HBase, Cassandra and such like.
All of them try to circumvalent slow HDD seek time by paralleling reads and writes across hosts in a cluster. Since SSD don't have this issue, and once these drives are durable enough old DB storages might be just scaled vertically. It will not last forever, by we might be able to ride this wave for another 10 years or so...

Igor

Thursday, November 12, 2009 6:09:59 AM (Pacific Standard Time, UTC-08:00)

That's a good point. Large memories and SSDs will absolutely help relational databases continue to be used successfully on a broad class of applications. Even more important is that is where the applications are and that's where the tools are. Old successful technologies never disappear fast and, in this case, there is no reason for relational systems to disappear at all. They are an excellent tool for many tasks. Just not the right tool for all tasks.

RDMBS aren't going away but they aren't going to be the right choice for all workloads either. I expect to see more database diversity going forward rather than less.

--jrh
jrh@mvdirona.com

James Hamilton

Comments are closed.