Good morning, we’re all feverish again with the well-known upgrade fever. But this time it is not our personal systems or the machines in the lab we’ll be upgrading but rather our server. As some of you know we’re running a co-located server with modest specifications that has handled traffic pretty well up until now. But we’re growing, and a bit faster than expected, so we need to spec up our server system to be able to handle the load.
Currently there’s a trusty old 440BX motherboard in our server running a Pentium III 800 at a 133MHz bus with 256Mb of PC-133 SDRAM. Storage is an Ultra160 10.000-RPM SCSI disk connected to an Adaptec 29160N SCSI controller. For local backup we have a 5400-RPM IDE-disk mounted inside the server that’ll do automated backups every 12-hours. We run RedHat Linux as an OS whereas the Apache webserver handles all of your HTTP requests, with HTTP compression that is.
That system will be upgraded to a 1.13GHz Pentium III and we’ll add 256Mb of PC-133 SDRAM in the process. We’ll also be moving up to a faster connection as we’re using more bandwidth than we anticipated and thus we’ll be connecting to the first router off of a 1Gigabits fibre connection directly on the AMS-IX. AMS-IX? We’re actually hosted in Europe and the AMS-IX is the fastest internet backbone available that connects directly to the transatlantic backbone. This system should be easily able to handle loads of up to 8-million hits per month, or +250K per day.
Although these numbers will surprise those of you that rely on Microsoft of other 3rd party server OSs I can only say that yes, Linux is that efficient, and has that little overhead. And let’s not start talking uptime shall we, our development server had more than 250-days of uptime even though we weren’t exactly gentle with it. The next step will be a separate database server connected to the first via a dedicated 1Gigabit Ethernet connection, again running Linux. That’ll probably have one, or two, fast 15.000-RPM Ultra160 SCSI disks for storage and again a 5400-RPM IDE-disk for local backups. We’ll be making this upgrade once we start to hit close to 8-million hits per month, but we’re not there yet, but we might be in the next few months.
Oh, just a little word on backups. As mentioned there’s an IDE-disk mounted in the server that does automated backups every 12-hours. It actually makes nice TAR-files out of the database and site content and stores them. Naturally we download those to our home machines every week, they’re not that big, just to be sure if a total system failure does occur. In case of a drive error on the 10.000-RPM SCSI disk we only need to insert a new drive with the pre-installed RedHat Linux OS and copy the database and content over and we’re good to go again. In the future we might opt for a redundant RAID1 SCSI solution, if budget permits, for now we’ll have to make do with a couple of hours downtime if something goes wrong.