New DB Servers
Our Dell PowerEdge 2950 Servers for the new DB Cluster has just arrived and I’m very excited, unpacking them. 
But before we start here a few specs:
- 2x Quad-Core Xeon E5440 2.8GHz/2×6MN CPU
- 32 GB 667 MHz RAM (8×4GB dual rank DIMMs)
- 6x 146 GB SAS (15′000 rpm) HDDs (RAID10 configuration)
- Integrated PERC 6/iR Raid Controller
- Dual Power Supply
- Dual Gigabit Ethernet
Looks Nice! Let’s see what’s inside
For the Motherboard enthusiasts, this one is for you
Quite well organized, but the interesting things are still covered… So I now open the shafts.
32 GB of RAM nicely lined up
After a few colleagues got attention to the open Hardware and a few minutes of watching so called “server porn” I lift up the Server. Time to make a front image (these six HDs are preconfigured as RAID 10):
After all, the first server now stands in our “testing room” and is ready for setup. In our case CentOS 5 will be running on it.
These servers will be our new DB-Cluster, which should improve DB performance significantly.
Experiences after the install and performance tests will follow…
Unpack porn! How great’s that!
Comment by
Leo Büttiker — 25. February 2008 @ 09:10
<–wants that too
looks nice!
Comment by mix — 25. February 2008 @ 10:36
Very nice - Those monsters should do a couple of thousand users…
Comment by Nitek — 25. February 2008 @ 12:02
@Nitek.. hope for millions users *gg*
Comment by esandar — 28. February 2008 @ 16:23
Ho my god, nice servers
I would like 10 servers PowerEdge 2950 me too for my work’s
Comment by Sammy.D — 28. February 2008 @ 16:46
Wow Pretty sweet! I wish I had them!
Comment by Ayerthon — 15. March 2008 @ 04:22
How do you guys decide when it’s time to upgrade your dataase servers? What is the key factor?
Comment by Ben — 17. March 2008 @ 23:08
Are you now going away from the “each-webserver-a-mysqldatabase” version of the site with these babys?!
Comment by ghost — 19. April 2008 @ 11:39
@Ben: When they are slow! Well we do monitor a lot of metrics in Ganglia and when we see that the host gets very high load during peek time and response not well anymore it’s time to call Dell!
@ghost: Hmm, yes that’s the plan (and we do it currently on some hosts). But this has not really something to do with this new babys, they are just a replacement for our masters that get a bit old now. In the near future we will seperate the webserver from the mysql slaves, a project that’s on our todo list for ages.
Comment by
Leo Büttiker — 20. April 2008 @ 19:15
Just to specify your answer Leo, the plan with this “babys” IS to get rid of the poor apache-mysql combination.
They will be run in a master-master setup (two pairs) with additional and separated hardware for some slave machines which will handle our SELECT queries (btw the ratio between SELECT and INSERT is 1:3 on our website).
We will run different databases on each pair (break up certain parts of the database and functionality to be able to scale better and faster).
Comment by Stefan Rothenhofer — 21. April 2008 @ 08:08