Re: > Subject: Re: Detailed explanation why uber move from postgress to mysql
- From: Mladen Gogala <gogala.mladen@xxxxxxxxx>
- To: Kellyn Pot'Vin-Gorman <dbakevlar@xxxxxxxxx>
- Date: Fri, 23 Dec 2016 22:54:29 -0500
On 12/23/2016 09:23 AM, Kellyn Pot'Vin-Gorman wrote:
Uhm, MySQL has shard query, sharding technology in MySQL is more
advanced than Oracles to begin with...) and Heap tables, MySQL's older
technology for memory tables has been around a significant amount of
time. It's not apple for apples, but depending on the use case, MySQL
has the features.
Kellyn
Hi Kellyn,
In memory technology maintains columnar store in memory, together with
row store. The nett effect is as if a bitmap index is created and
maintained on the columns, minus the locking problem. This advanced
algorithm can speed up aggregated functions like avg, sum or stddev
functions tremendously. I have been using ENGINE=MEMORY tables a long
time ago, by creating a "sales" database from Oracle on the nightly
basis. Oracle was doing the heavy "group by" lifting and the results
were inserted into a MySQL database using Perl scripts. Sales people
were then accessing it using Crystal Reports and Business Objects, two
tools that used to create atrocious queries against an Oracle database.
The results were flying. I don't, however, think that this could compete
with the in-memory technology employed by Oracle and other advanced
commercial databases. The first technology of that type was "BLU
acceleration" in DB2, a year older than Oracle's own in-memory option.
Interestingly enough, MySQL is entering the fray:
http://www.infoworld.com/article/3150683/analytics/mariadb-crashes-open-souce-big-data-analytics-competitors.html
MariaDB, an open source variety of MySQL has obviously developed
something very similar to Oracle's in-memory technology. Now, that would
be an excellent reason for moving from Postgres to MySQL.
--
Mladen Gogala
Oracle DBA
http://mgogala.freehostia.com
Other related posts: