One of the things I’ve noticed (and I’m sure you have too) with Server 2008 is that we still see many scenarios in which a reboot is required. For example, adding or removing many of the roles in Server Manager requires a restart. Microsoft has tried a number of tricks (see: delayed-start services) to make these restarts faster, or at least to make them feel faster, but it still takes a pretty long time on many machines before they are fully functional. And don’t forget those reboots we perform in troubleshooting sessions that don’t seem to respond to anything else!
Server 2008 is light-years ahead of Windows NT in the “number of times you have to reboot” department, but I haven’t seen a big improvement compared to Server 2003, which makes me wonder whether something fundamental in the Windows architecture dictates that we’re going to be rebooting our systems more than we’d like.
So, what can we do? One option that’s been around for at least a couple of decades is solid-state disks (SSD’s). These have no moving parts and use nonvolatile memory to emulate traditional spinning-disk storage functionality. Surf the net for a while and you will see dozens of speed tests demonstrating that using an SSD for your system disk can cut boot times in half. That’s pretty significant, especially with mission-critical servers.
SSD technology doesn’t have the same issues as spinning disk technology, and there are some “gotchas” to be aware of. We’ll take a look at some of these in upcoming columns. And SSD’s are a LOT more expensive than traditional spinning disks, although SSD prices have been dropping fast in recent years. But you could set up servers so that only the system disk lives on SSD, using spinning disks for data volumes, and spend comparatively little money for a quantum jump improvement in boot times (and, potentially, application load times too). That’s very interesting indeed. Stay tuned for more discussion of this topic.
