Cool, but why?
deleted by creator
I’m curious, does a 3 minutes power down to replace a RAM stick is that much of a deal in enterprise server that they need to invented a whole new technology just for that?
The hangup is that you think shutting down and restarting a server takes 3 minutes
The surplus enterprise hardware I have in my homelab takes 3 minutes to just get to BIOS
deleted by creator
yeah.
The coveted 5 9s of availability is only 5.26 minutes of downtime
5.26 minutes per year*
Yes. Server boot times are long. Enterprise level NICs and hard drive controllers do a lot of checking at startup.
Historically, there were Sun servers that could hot swap CPUs. X86 can’t do that, though.
Many that weren’t based on x86 microcompters could do this: Tandem, I mean, Compaq, I mean HP NonStop machines, Sun Ultra Enterprise as you mentioned, IBM s390 and System-Z, several HPUX systems, I’m sure there’s others.
First of all, yeah. In enterprise, 1000 transactions per second can be a requirement. Second, enterprise servers take longer to spool up than 3 minutes.
Depending on your SLA, 3 minutes can be a pretty big chunk of your monthly error budget.
Have you ever power cycled a server? It can take over 10 minutes depending on the machine.
Meanwhile, Windows requires you to buy a new license if you change your mouse.
That’s not entirely true. You can get a universal mouse license that will cover up to twenty approved mouse changes.
Did anyone else notice that in their stock photo they’re trying to put the DIMM into the socket backwards?
You want that distinct crunch as you force the RAM in with a mallet
same with the CPU, the last thing you want are straight pins yuck
Every stock photos is like a mini Easter eggs hunt.
I don’t understand how this is an advantage. Yes, you can swap RAM with the system powered up, but what happens to the information in the module that was removed? Is the OS doing some kind of RAID-like memory allocation? The article wasn’t clear on how this would actually work.
Servers have had memory mirroring as a feature for years. This seems like a cool extension of that technology. It would be an advantage in some systems where scaling out isn’t an option and single node availability needs to be as high as possible.
Apparently, there’s some coordination mechanism, where you tell the OS that you want to remove a certain memory stick, so it moves all the memory onto other RAM sticks (or uses paging to move it to your hard drive). Only then would you actually physically unplug the memory stick.
See, for example: https://docs.kernel.org/admin-guide/mm/memory-hotplug.html
(Mind that this is kernel documentation. If you actually want to do this, there’s probably some CLI program to make it easier.)Then it’s not Hot Swap, just Lukewarm Swap?
I guess, you could see it that way…? The important part is that you don’t have to turn off the whole system. It can continue running without interruption. So, the RAM will be lukewarm when you swap it, but the system will still be hot.
If you really want to be pedantic you could setup raid 1+0 or 5 and live the true RAM hot swapping life
So “warm plugging” is a thing - it means a piece of hardware is detachable while the machine is asleep.
I remember the ‘good old days’ of Sun Fire 10k and similar servers. You could replace entire boards of CPU and RAM and the server would keep on trucking.
IBM System-Z mainframes still support CPU hot-swapping AFAIK.
Don’t. I still miss Sun. I miss that “Hey, fuck MS Office’s crazy pricing. Let’s buy their competitor for less and open source it!” energy.
We need more of that in these times.
I remap Caps Lock to Ctrl to this day as that’s where it’s meant to be, god damn it!
Excellent! Hot swap all the things!!!
FInally, Banjo Tooie on PC