If you want a more modern VME bus PC, then go take a look at CompactPCI hardware -- it's basically PCI in a VME-like formfactor. Ziatech has a whole pile of P2 boards, as do Moto rola and quite a few other vendors. They aren't cheap, though. If we could get cPCI hardware for less than a 100% price premium, I'd be seriously considering them in a lot of places that we're using more traditional rack-mount PCs today.
It's actually pretty easy to support this, if you have the right hardware. I support roughly 100 Linux/x86 servers spread across the US, and hardware makes a big difference.
First, by and large, remote rebooting doesn't matter as much as you might expect. Over the last year, we've only had 5 or so actual lockups in remote locations. Generally, under Linux, when something dies, it's dead, with a total hardware failure. For the most part, power supplies and hard drives are to blame, although we've lost a motherboard or two. Avoid cheap drives and power supplies -- Seagate Barracudas work almost flawlessly for us. Frankly, we've had more problems with spontanous reboots that hard lockups.
If you're paranoid (we are:-), and want to do things right, then the first thing to look at are Intel's server boards. We use the T440BX and the N440BX (single/dual P2 boards). They both support serial consoles in the BIOS, they have built in SCSI, Ethernet, and VGA, they have a on-board hardware watchdog, they have excellent hardware monitoring built-in, and they support remote reboot/power-off/power-on, if you can figure out the protocol to use. I have a driver for 2.2 that supports quite a few of the monitoring/management features of the board. Mail me for info.
Slashdot Mirror
If you want a more modern VME bus PC, then go take a look at CompactPCI hardware -- it's basically PCI in a VME-like formfactor. Ziatech has a whole pile of P2 boards, as do Moto rola and quite a few other vendors. They aren't cheap, though. If we could get cPCI hardware for less than a 100% price premium, I'd be seriously considering them in a lot of places that we're using more traditional rack-mount PCs today.
It's actually pretty easy to support this, if you have the right hardware. I support roughly 100 Linux/x86 servers spread across the US, and hardware makes a big difference.
First, by and large, remote rebooting doesn't matter as much as you might expect. Over the last year, we've only had 5 or so actual lockups in remote locations. Generally, under Linux, when something dies, it's dead, with a total hardware failure. For the most part, power supplies and hard drives are to blame, although we've lost a motherboard or two. Avoid cheap drives and power supplies -- Seagate Barracudas work almost flawlessly for us. Frankly, we've had more problems with spontanous reboots that hard lockups.
If you're paranoid (we are :-), and want to do things right, then the first thing to look at are Intel's server boards. We use the T440BX and the N440BX (single/dual P2 boards). They both support serial consoles in the BIOS, they have built in SCSI, Ethernet, and VGA, they have a on-board hardware watchdog, they have excellent hardware monitoring built-in, and they support remote reboot/power-off/power-on, if you can figure out the protocol to use. I have a driver for 2.2 that supports quite a few of the monitoring/management features of the board. Mail me for info.
Scott