Comments for Dan Weinreb's blog

Comment on Comments on “Urban Myths about NoSQL” by Dan Weinreb

Dan Weinreb — Sun, 10 Jul 2011 14:13:09 +0000

Oops! Thank you very much for catching that; I’ll fix it.

Comment on Comments on “Urban Myths about NoSQL” by Christian G. Warden

Christian G. Warden — Fri, 08 Jul 2011 16:40:16 +0000

Under, #5, you mean “network connection will have latency so *high* that…”, right?

Comment on Comments on “Urban Myths about NoSQL” by Dan Weinreb

Dan Weinreb — Fri, 24 Jun 2011 16:00:34 +0000

Hi, Fred. Yes, by “scale” I did indeed mean the number of servers. That’s usually how that word is used in this kind of context. However, if it was misleading, I apologize. Yes, if you can get more work accomplished with fewer nodes, then the problem of network failure goes down: good point. Thanks!

Comment on Comments on “Urban Myths about NoSQL” by Fred Holahan

Fred Holahan — Fri, 24 Jun 2011 14:42:59 +0000

Hi Dan, I remember reading a post like this from you a while back and, so, am pleasantly surprised to see that it appears to have been posted again recently.

I’d like to elaborate on your point 6, in which you say “My impression (I may be wrong) is the the ‘sweet spot for VoltDB, at least at the moment, is for distributed systems that are not at the kind of very-large scale of an Amazon or Google, and indeed for a much smaller scale, which makes network partitions much less of a problem.”

Your characterization of scale there is a bit misleading. If “scale” were only to mean “number of nodes in a database cluster”, I’d agree with your statement. On the other hand, if “scale” means “the amount of mixed-workload throughput the database cluster can deliver” (by my experience, this is what most IT users really care about), then your characterization is a bit misleading.

To say this another way, VoltDB doesn’t need a massive number of nodes to deliver eye-popping database throughput. You can achieve hundreds of thousands to millions of ACID TPS on a relatively small cluster (at K=1, 16 nodes is a pretty big VoltDB cluster). To your point, smaller numbers of nodes generally translate to lower probability of network partitions. But this is a strength for Volt, not a limitation. Since Volt’s per-node throughput delivers excellent scale on relatively small clusters, the risk of network partitions is significantly lower with Volt than with scale-out alternatives that require many more nodes to achieve comparable throughput.

I’d say that the “sweet spot” for VoltDB is applications that have partitionable datasets and run over a fast network on a relatively small number (i.e., tens not hundreds) of database nodes. In such configurations, Volt can reliably deliver very favorable performance and scale.

Comment on VoltDB versus NoSQL by Tom G

Tom G — Thu, 05 May 2011 19:04:53 +0000

Additional problem with ad-hoc – code injection. If it can be defined and executed, it’s a problem.

Oracle has had this issue with PL/SQL since it was introduced.

Comment on Errors in Database Systems Still Must Consider Network Partitions by Dan Weinreb

Dan Weinreb — Mon, 18 Apr 2011 13:35:33 +0000

Alex Feinberg, a comment to another blog item herein, referenced this excellent paper, which made many of the same points I’m making. Had I known about it at the time, I would have cited it. It’s very well-written! It’s from Henry Robinson, posted on the Cloudera blog:

http://www.cloudera.com/blog/2010/04/cap-confusion-problems-with-partition-tolerance/

Comment on Why Did M.I.T. Switch from Scheme to Python? by Feedback Loops in Software Development – Irrational Exuberance | Agile Development

Fri, 01 Apr 2011 14:41:02 +0000

[...] you believe MIT’s switch from Scheme to Python is a concession or a catastrophe, I don’t think it is a [...]

Comment on VoltDB versus NoSQL by ehcache.net

ehcache.net — Mon, 14 Mar 2011 13:11:03 +0000

VoltDB versus NoSQL…

Mike Stonebraker is the co-founder and CTO of VoltDB, which makes a novel on-line transaction processing (OLTP) relational database management system (RDBMS). He recently gave a talk entitled “VoltDB Decapitates Six SQL Urban Myths”. You can read the s…

Comment on Boston: No More Silicon Valley Envy! by Dan Weinreb

Dan Weinreb — Sat, 12 Mar 2011 14:09:30 +0000

@Terry: This is about non-competes being null and void in California. I discussed this in an earlier comment. I hope we can get this in Massachusetts!

Comment on Boston: No More Silicon Valley Envy! by Terry

Terry — Sat, 12 Mar 2011 08:57:44 +0000

There is a reason why Silicon Valley can be in CA only:

http://www.netvalley.com/silicon_valley/Legal_Bridge_From_El_Dorado_to_Silicon_Valley.html