Tuesday, 24 April, 2018
Last year we already published our over 2.1M
record with MySQL 8.0 -- it was not yet GA on that moment and the
result was obtained on the server with 4CPU Sockets (4S) Intel Broadwell
v4. We did not plan any improvement in 8.0 for RO related workloads, and
the main target of this test was to ensure there is NO regressions in
the results (yet) comparing to MySQL 5.7 (where the main RO improvements
were delivered). While for MySQL 8.0 we mostly focused our efforts on
lagging WRITE performance in MySQL/InnoDB, and our "target HW" was 2CPU
Sockets servers (2S) -- which is probably the most widely used HW
configuration for todays MySQL Server deployments..
However, not only SW, but also HW is progressing quickly these days ! -- and one of my biggest surprises last time was about Intel Skylake CPU ;-)) -- the following graph is reflecting the difference between similar 2S servers, where one is having the "old" 44cores-HT Broadwell v4, and another the "new" 48cores-HT Skylake CPUs :
the difference is really impressive, specially when you see that just on 32 users load (when CPU is not at all saturated not on 44cores nor 48cores) there is already 50% gain with Skylake ! (and this is about a pure "response time"), and on peak QPS level it's over 1.8M QPS (not far from 80% gain over Brodawell)..
And this results is marking our next milestone in MySQL RO performance on 2S HW ! ;-))
Sysbench RO Point-Selects 10Mx8-tables latin1 on 2S 48cores-HT Skylake
As already mentioned, the main gain is coming from MySQL 5.7 changes, and we're probably just little bit lucky here to see MySQL 8.0 slightly better than 5.7 ;-)) (while as you can see from the chart, it was also a good reason for MariaDB to move to InnoDB 5.7 to match similar gains comparing to InnoDB 5.6)..
So, it's as expected, to not see any difference on mixed OLTP_RO :
Sysbench OLTP_RO 10Mx8-tables latin1 on 2S 48cores-HT Skylake
and what is amazing here is that we're reaching on 2S Skylake now the 1M QPS result that we obtained in the past with MySQL 5.7 on the same 4S Broadwell v4 box (which was not yet upgraded to 96cores on that time).. And it makes me smile now to recall all the discussions with our users mentioning "they will never use anything bigger than 2S server" -- and here we're ! -- its exactly 2S, 48cores "only" box, but pointing on ALL the problems we already investigated on bigger HW and fixed them ON TIME ! ;-)) -- well, there is still a lot of work ahead, so let's hope we'll be "always on time", let's see..
The most "sensible" RO workload for me was always "distinct ranges" in Sysbench, as it's pointing on issues which anyone can have with "group by" or "distinct" queries.. -- which is involving temp memory tables code, and this code was completely rewritten in 8.0 "from scratch" to be finally more simple, "maintainable" and more scalable. But this is not always meaning "as efficient as before" (the most efficient it could be probably if it was re-written on assembler, but again -- who will want to maintain it again ? ;-))
Sysbench RO Distinct-Ranges 10Mx8-tables latin1 on 2S 48cores-HT Skylake
so far, MySQL 8.0 is doing mostly the same as 5.7 here, and it's really good keeping in mind the impact of the changes.. (while further improvements may still be done, as well many old "broken" cases are solved now, etc.)
(but have nothing to say about MariaDB 10.3.5 here)
well, this was about "latin1" RO results, stay tuned, there is more to come ;-))
And Thank You for using MySQL !
P.S. an "attentive reader" may ask himself -- if over 1.8M QPS are reached on 2S Skylake, what QPS level then will be matched on 4S Skylake ??? -- and I'd say you : we don't care ;-)) we know the result will be higher, but we're NOT running after numbers, and there are other more important problems waiting on us to be resolved, rather to show you "yet more good numbers" ;-))
blog comments powered by DisqusNote: if you don't see any "comment" dialog above, try to access this page with another web browser, or google for known issues on your browser and DISQUS..