Server ORANGE upgrade

If I could pick one thing to do all day long for the next 10 years, without a moment’s hesitation, it would be tinker with server hardware. I’m talking about the kind of stuff that your IT department geeks are too ashamed of talking about by the water cooler.

Unfortunately (or fortunately) my role in the company has drifted into another arena and I rarely get to geek out. Yesterday (Wednesday, Nov 1st) was an exception. Shared server ORANGE, is a mature server that has been running for almost two years without any major issues. It hosts several hundred websites and is by no means overloaded. But in the past two weeks it was having high server load that could not be explained.
While only a handful of customers noticed that email was arriving slower than usual, our network admin was busy putting out this “fire” by following every procedure to decrease load. This helped stabilize the situation somewhat, but it was a patch at best, and by no means a solution.

When all possibilities were exhausted, I was called in to make the final call. My directive was to upgrade the hardware and take our users out of the line of fire, and then troubleshoot the old parts to see what is causing the problem.  To make the story a little shorter, that’s exactly what they did and it made a tremendous difference.
So if you ever wondered what some of these systems consist of, here is a base line shared hosting system that we use today:

• 2 Dual Core XEON 5130 (2.00 GHz) Woodcrest CPU – 4 Processors total

• 1366 MHz bus speed

• ECC Registered RAM

• Hardware RAID Controller for SATA Drives

• WD Raptor 1500ADFD hard drives with 10,000 RPM drive speed fastest in its class

Normally the shared servers use SCSI hard drives which are much faster than SATA, but to ease the transition of an already working server, WD Raptors were used because we could simply take out one of the old drives and rebuild the array onto the new (much faster) Raptors, one at a time.

It sure worked well! No delays of any sort!  And it turned out to be a bad drive in the array that could not keep up with the rest of the healthy hard drives and was slowing down write operations.
So the moral of the story, most problems disappear if you just throw money at them.

Artur November 2, 2006