Comments on: MySQL and hardware information

By: Sheeri

Sheeri — Wed, 18 Nov 2009 16:15:09 +0000

Camila — it’s using regular linux-ha.

By: Camila

Camila — Tue, 17 Nov 2009 20:14:48 +0000

Sheeri
you said “The primary site has 2 machines connected to the same LUNs on a HA
setup”. Could you say more details about this ha setup? ex, is it linux ha ou custom scripts…
Thanks!

By: Matt Reid

Matt Reid — Sat, 31 Oct 2009 19:59:49 +0000

Here are some real-world stats from some of the servers I admin. There are more big ones but these are just the ones on the top of my head right now. MySQL scales well and pushes tons of traffic on the internet – you just have to design it to do so correctly.

1. 50K queries per second, 10K connections – internet ad serving company. several slave servers for reporting and analytics. hardware: 8 core opterons, 32GB ram, attached to an EMC SAN utilizing 120GB dataset. many small queries and query cache hits. several scripts running every hour to ensure the query cache is preloaded with the proper data. MySQL 5.0.68-percona

2. 8K queries per second, 1000 connections, lots of memcached and custom caching on the app – also an internet ad serving company. several slave servers for reporting and analyics. hardware: 16 core, 64G ram, EMC SAN using 180GB dataset. MySQL 5.1.x -percona

3. 2K queries per second, 1400 connections, 32 CDN servers + 4 java bytecode processing servers, 4 slave servers for read-only traffic. hardware: 8 core opterons, 32GB RAM, EMC SAN with 250GB data space. MySQL 5.0.68-percona

By: Log Buffer #167: a Carnival of the Vanities for DBAs | Pythian Group Blog

Log Buffer #167: a Carnival of the Vanities for DBAs | Pythian Group Blog — Fri, 30 Oct 2009 17:04:54 +0000

[...] On the Pythian blog, Sheeri Cabral shared a brief survey of large data sets and their hardware stats in MySQL and hardware information. [...]

By: kedar

kedar — Tue, 27 Oct 2009 04:46:14 +0000

How about RAID or is that out of scope?
Long back I happened to be here [http://www.sizinglounge.com/index.php?swid[]=10&users[]=400&swid[]=10&users[]=400&appscount=2&aid=1&pid=1&util=70&storageSize=576&months=36&pue=25&discount=0&energy=16&labour=6000&license=0&vendors[]=10&cat[]=1&cat[]=2&cat[]=3&pf[]=1&pf[]=2&pf[]=3&pf[]=4&pf[]=5&submitted=TRUE]

How much does it make sense? Can any one explain considering default settings eg. my-large.cnf?

By: Gerry Narvaja

Gerry Narvaja — Mon, 26 Oct 2009 22:06:57 +0000

I have been working with databases for more years than Pythian has been around, and I have been asked this question too many times in my career.

Here Sheeri is presenting a few examples from the real world.

What I don’t really understand is *why* does it have 2 votes against in Planet MySQL, can someone explain this to me?

My $.02

By: Daniel

Daniel — Mon, 26 Oct 2009 21:50:16 +0000

We run a 2TB MySQL instance:

IBM HS21 blade server with 16GB RAM and 2 quad core Xeon cpus

Storage is an IBM DS3400 FC array with 24 15k RPM SAS disks

OS is OpenSolaris 2009.06

By: Mark R

Mark R — Mon, 26 Oct 2009 21:45:26 +0000

Quite interesting…

We also use Dell Powervault MD3000 with Dell 1950s with 32G of ram.

Our servers get almost *no* select queries (only 100s per hour at busy times), but a very high load of inserts, typically 300 rows per second 24/7 (not individual inserts; they insert in batches). Our tables have lots of columns and indexes; it’s an auditing reporting system.

We use MyISAM which does have some drawbacks but is very space efficient. We use PACK_KEYS and DELAY_KEY_WRITE as well as various other tweaks.

redundancy is got by using replication and slave servers with similar hardware but bigger discs; the master servers have typically 300G – 1.5T of data each.

By: Ryan

Ryan — Mon, 26 Oct 2009 21:09:45 +0000

For an I/O bound MySQL workload, I really like the SunFire X4140/X4150 — 8 local disks (with SATA/SAS/SSD options) in a 1RU box with 12 cores and 128GB of RAM and hardware BBU raid card.