4 ms·

I'd be interested to see more benchmarks. The improvements in this post are not anywhere close to what we've seen going from PG to Infobright for our reporting

by mason55 13y ago

I'd be interested to see more benchmarks. The improvements in this post are not anywhere close to what we've seen going from PG to Infobright for our reporting queries - we get speedups from 10x - 1000x, the one speed benchmark they have here is only 2x.

ozgune 13y ago
(Ozgun from Citus Data here.) We're going to publish more comprehensive results in a few weeks. In our initial tests, we found the speed up to depend a lot on the underlying data and the type of queries. For example, we found that when compression reduced the working set from being on disk to in-memory, there was a significant jump. Also, we're looking to do optimizations on the cost estimation side -- these will notably help with queries that join multiple tables together, a common scenario for the TPC-H benchmark.
- mason55 13y ago
  Awesome, can't wait to see more benchmarks. I would love to be able to switch back to 100% Postgres. IB gives us tons of speed but even ignoring the MySQL warts it just feels incomplete. And we ran into a bug in production where the order of AND was actually having an effect on query results, something that's completely unacceptable and makes me worry about my data.
alecco 13y ago
Because they still use PostgreSQL engine, very slow. There is a great talk by Stonebraker about this. (He created PostgreSQL about 3 decades ago and moved on to newer database engines).
- Alex3917 13y ago
  Is this talk online?
  - garenp 13y ago
    Probably referring to this: http://blog.jooq.org/2013/08/24/mit-prof-michael-stonebraker-the-traditional-rdbms-wisdom-is-all-wrong/ http://blog.jooq.org/2013/08/24/mit-prof-michael-stonebraker...
- jeffdavis 13y ago
  He started Postgres before it became PostgreSQL (i.e. before it even supported SQL). It's quite a different system now.
- jeltz 13y ago
  The PostgreSQL Stonebraker left was a very different one than the current. It did not even have MVCC.
antonmks 13y ago
Doesn't look like it uses any kind of vectorized execution. So the only advantage one gets is reduced amount of data to process. It also doesn't mention execution on compressed data so it has to decompress the data which doesn't come cheap.
- erichocean 13y ago
  A relatively easy way to change that would be to use Intel's ISPC compiler. http://ispc.github.io/ http://ispc.github.io/

by mason55 13y ago

ozgune 13y ago

(Ozgun from Citus Data here.) We're going to publish more comprehensive results in a few weeks. In our initial tests, we found the speed up to depend a lot on the underlying data and the type of queries. For example, we found that when compression reduced the working set from being on disk to in-memory, there was a significant jump. Also, we're looking to do optimizations on the cost estimation side -- these will notably help with queries that join multiple tables together, a common scenario for the TPC-H benchmark.

mason55 13y ago
Awesome, can't wait to see more benchmarks. I would love to be able to switch back to 100% Postgres. IB gives us tons of speed but even ignoring the MySQL warts it just feels incomplete. And we ran into a bug in production where the order of AND was actually having an effect on query results, something that's completely unacceptable and makes me worry about my data.

alecco 13y ago

Because they still use PostgreSQL engine, very slow. There is a great talk by Stonebraker about this. (He created PostgreSQL about 3 decades ago and moved on to newer database engines).

Alex3917 13y ago
Is this talk online?
- garenp 13y ago
  Probably referring to this: http://blog.jooq.org/2013/08/24/mit-prof-michael-stonebraker-the-traditional-rdbms-wisdom-is-all-wrong/ http://blog.jooq.org/2013/08/24/mit-prof-michael-stonebraker...
jeffdavis 13y ago
He started Postgres before it became PostgreSQL (i.e. before it even supported SQL). It's quite a different system now.
jeltz 13y ago
The PostgreSQL Stonebraker left was a very different one than the current. It did not even have MVCC.

antonmks 13y ago

Doesn't look like it uses any kind of vectorized execution. So the only advantage one gets is reduced amount of data to process. It also doesn't mention execution on compressed data so it has to decompress the data which doesn't come cheap.

erichocean 13y ago
A relatively easy way to change that would be to use Intel's ISPC compiler. http://ispc.github.io/ http://ispc.github.io/