== How old our servers are (as of 2022) Back in 2016, I wrote about [[how old our servers were at the time OurServerAges2016]]. They were rather older than people might have expected, because universities are generally cheap and so usually run servers much longer than many people do. [[My group https://support.cs.toronto.edu/]] no longer quite runs servers into the ground, but we still can come close. Today, for reasons beyond the scope of this entry, I'm going to do a 2022 version of [[my old entry OurServerAges2016]]. [[My group]] only handles general departmental infrastructure on the non-undergraduate side of things, although these days we have some [[big servers ../linux/BigServerFastFirefoxBuild]] that are mostly in [[our compute cluster SlurmHowWeUseIt]]. However, most of the most modern and powerful servers are in research groups, and get turned over much faster than we do (in fact we just recently got rid of some vintage 2011 'compute' servers we inherited that way). Our normal servers remain almost entirely 1U Dell servers, although we've wound up with some ultra-short Supermicro servers as well that we use for firewalls. What we consider our current generation of Dell 1U servers are R340s and R240s; these are what we use for new installs of machines that we particularly care about. Since we're in the process of upgrading a bunch of machines from Ubuntu 18.04 to 22.04, the number of these servers in production use is likely to go up. Somewhat older than that are Dell R230s, which it looks like we started using in 2017 or maybe 2018, and then we have quite a number of R210 IIs and R310s still in service, although we're rotating those out of service as we upgrade machines to Ubuntu 22.04. We're still reusing some of these old Dells for test servers or unimportant things, although we've decided that a number of them have CPUs that are now just too slow for modern Linux. (Linux can run on even relatively slow machines, but things like installing Ubuntu 22.04 can be painfully slow. When we have better hardware, the hassle isn't worth it.) In [[our compute cluster]] we have a bunch of Dell 3930s with GPUs, a few [[modern fast Supermicro servers ../linux/BigServerFastFirefoxBuild]], a bunch of hand-built machines using ASRock X399 Taichi motherboards and AMD Threadrippers, and some assorted miscellaneous machines. Until earlier this summer, we were still running a number of donated Dell C6220 blades in the compute cluster, but with the addition of some modern Supermicro servers we finally felt that we could eliminate the now very old and not very impressive C6220 blades. [[Our Linux fileservers ../linux/ZFSFileserverSetupIII]] remain on their vintage 2018 Supermicro hardware with very low odds of that changing. Probably this is fine, especially with the relatively slow advance of basic server performance these days. We've very recently reused some of [[the older fileserver hardware ../solaris/ZFSFileserverSetupII]] for a new purpose; it's not very fast, but we have lots of spares and it can hold plenty of disks. Conveniently, Dell lets you look up the age of machines online if you have their service tag. This provides eye-raising numbers for our R210 IIs and pretty alarming ones for the few R310s that we have in service; for example, [[our central syslog server CentralizeSyslog]] turns out to be on hardware that's about a decade old. Since it's running CentOS 7, we're probably still not going to touch it for another year or two. (Some of our hardware turnover since 2016 was motivated by [[the CPU vulnerabilities discovered since then and their effects on old servers IntelMDSKillsOldServers]].)