Hi everyone. I would like to take this opportunity to thank all our speakers for their great presentations. As the time for individual presentations was limited and you may have some additional questions to our presenters, we would like to give you an opportunity to learn more about the companies and the products presented during our event. Ori Herrnstadt, the founder and CTO of Akiban, is organizing a WebEx webinar NewSQL follow up: How to get document style access with the full power of SQL.
0 · May 3, 2012
@Stefan: it's pretty much as you've observed: it's good for one set of joins, but not for another. That is, I can join multiple tables on a single key, but I can't model transitive relationships of the kind t1.a = t2.a and t2.b = t3.b and .... It's just the nature of the beast: if you're going to physically organize the data to support a particular query, it supports that particular query. Unless you materialize replicas organized in other ways, you only benefit that particular query.
0 · April 22, 2012
While I certainly appreciate effort to make this happen, it was poorly run. There appeared to be no attempt to keep the presenters within the 5-10 minutes they were allotted. Because of that some presenters went way overtime and that was not fair to the other presenters or the audience. And there was no time left for discussion or networking at the end. The method of projection of the presentations was poorly set up. And there was absolutely no need for a microphone and amplifier in a room that small and I felt its use was actually detrimental.
0 · April 19, 2012
Hi Chris, glad you spoke up. I was wondering during the meeting if this was new or not. What I would be interestined in is this: if there are multiple queries that all want to pre-join data in different ways, is there some way to do this without picking losers and winners? Otherwise one is organizing the data to serve a fixed set of queries, not that there is anything inherently wrong with that, but choice is even better.
0 · April 18, 2012
@Stefan Gower: re Akiban: Oracle has been doing this with what they call "clustered tables" for over 25 years. Member tables of a "cluster" (not a physical device concept here) share storage space, and all records with the same "cluster key" are stored together, effectively pre-joining them. As you observe, it would be good to know for all these solutions what the situations are that they really help in (like this case) in more detail.
0 · April 17, 2012
It was an interesting meeting, but it was a shame that there was no Q&A because some speakers overran their allotted time in the lightning presentations. But the speakers did a great job.
Akiban was very interesting, but in a conversation with others, our little group all felt that the approach was a very targeted approach to handling joins. While such an approach could be *highly* advantageous in certain applications, nothing in the presentation addressed when the Akiban solution would be advised, and when it would not be. None of these remarks should be interpreted as being negative on Akiban, but just a desire, as with all solutions, to understand the application conditions where the solution is best applied, and where it breaks down. Such an understanding of these breakdowns might even suggest further ways to generalize the technology and thus make it more widely applicable.
0 · April 17, 2012