My new server is a member of Dell’s 12th server generation. A lot has changed over the years. 20 or 25 years ago, servers were pretty much the same as desktops. Well, they were the same as tower desktops, which usually spent their life on the floor under the desk. The first innovation that I remember was adding racking ears and twisting the logo ninety degrees so that it would read right when the machine was on its side. There was some controversy about whether the disk drives would be less reliable if they spun on a horizontal axis (they weren’t).
Over the years, Intel developed processors aimed at the server market, only to find that they were adopted by the manufacturers of high-performance workstations, so that was only briefly a distinction.
After going through the spec-ing process and working with my new server for a while, I think there has been significant divergence from the desktop market.
Form factor. The new servers are rack mountable, just like the old ones, but they are designed to save on vertical rack space, and are quite deep. My new server is 2U high (3.5 inches), and has room for 8 hard disks, two power supplies, and a big motherboard.
Redundancy. The new server has redundant power supplies, redundant cooling fans, redundant (mirrored) SSDs, and even redundancy in the memory system. There’s ECC, of course, but it doesn’t stop there. If a rank of a memory module fails, the motherboard will swap in a rank from another module. This is apparently supported by the OS, since it shows 32GB installed and 24GB available. I imagine that it would take a restart to recover from a complete rank failure. I am not in a hurry to find out.
Remote operation. Modern servers are designed to be operated from a distance. In addition to various software ways of making this happen on a fully functioning system, which could work on any hardware, the Dell remote management hardware allows the system to be powered up remotely (powering down is easy), and allows the BIOS to be configured remotely. You can do anything remotely that you could do at the server that doesn’t involve messing with wires or taking the covers off.
Uptime maximization. Not only the disks are hot swappable, so are the power supplies.
Identification. Not that I have the problem, but, if you have a lot of servers, when you do have to actually mess with the hardware, it can be a chore to find the right one. Being able to display a wide variety of information on the front panel of the computer (including, should it be desired, “This one!”) can certainly help there. When you’re working from the back of the rack, it’s good to know when the system’s powered up; the server has a LED on its rear panel just for that purpose.
Scale. Here’s a screen shot of the remote management console’s computer location section:
Leave a Reply