IT projects can arise from the most interesting of circumstances. One project begets another project and another and down the rabbit hole you go. The subject of this article is power management; a project brought to the forefront by of all things our upcoming replacement of desktops with low-power Wyse thin clients. If that …
VDI on ESX with HP iLO
With VMWare building DPM into its systems VDI on ESX with HP kit, DPM connected to iLO and the whole lot in a automated DRS cluster means we can have servers power up and down according to load, and rebalance the cluster as needed. Users don't see any performance slowdown thanks to PAM cards on the NetApps we use for storage and we save on power overnight and weekends as most of the hosts power down.
"Based off of"
A house is built UPON its foundations.
Things are not based "off of", or even based around.
An object having a base sits ON that base.
You fail because you do not consider the meaning of the words you use to communicate.
This sort of thing makes me remember that wintendo is complete and utter crap with a cherry on top. Yes, it performs quite nicely. No, it's not manageable in any halfway readonably straight-forward fashion, unless you're sitting in front of it and you accept screen, keyboard, and often mouse, as normal prerequisites for a server.
Me, I think Real Servers Are Headless. Headless Are Real Servers. You know, serial console, tack on a modem or a (reverse) terminal server, off you go. Have a FEP to control the power. That sort of thing.
This is of course entirely alien to the GUI-weaned crowd. But it eats a lot less bandwidth and keeps on working in much more adverse conditions. Me, old, bitter, caricature of myself? You have to ask? But still, serial ports do have merit, moreso than USB.
Oh, did you perchance measure how much power the machines want when off? WOL isn't entirely free, and it is worth knowing what the cost is.
IPKVM is good for my ESXi and Linux servers too. There are many things that can go wrong before the OS loads, and before that happens, all operating systems are equal.
Also: 18 watts on average. They run around 450 W idle and 750 W loaded. Idle would be much better if they had 80+ PSUs
Give it some head...
So "Real Servers Are Headless", well apart from the serial terminal and the FEP controller and don't forget the serial-connected printer you'll need so you can work through the five thousand lines of log data with a highlighter to find out why the OS blew out last time as opposed to scanning a GUI-driven editor with context-sensitive folding and search options... Still it all makes work for a working man to do, I suppose.
"and don't forget the serial-connected printer you'll need so you can work through the five thousand lines of log data with a highlighter to find out why the OS blew out last time as opposed to scanning a GUI-driven editor "
Odd. Most of the log files I've seem have pretty simple formatting (CSV or even TXT) which *can* be read by something as primitive as edit in a DOS box.
BTW Dumb terminal editors *can* come with very effective folding line and search/replace functions. I'm old enough to have used some good ones and I consider it ridiculous that something called a *server* should whine that it's got no keyboard/mouse/screen attached to it.
Anyone remember WordStar?
Also: wake-on-LAN for upgrades... but maybe you don't want to wake every PC in the building at once, for that...
...then again, on a cold day start booting and updating PCs remotely at 7am, and by the time folks get to their desks the office will be nice an' warm...
.. did you say that your fat-desktop fleet is end-of-life AND expanding? Life's cruel. :-)
About half my fat-desktop fleet are ASUS P5E-VM-DOs, (vPro boards) with C2Qs and 4GB of RAM. Perfect little Photoshop boxes. The other half are ASUS K8N-DLs with dual Opteron 940s and 4-6GB of RAM.
The K8N-DLs were our original fleet of Virtual Servers. They served out their warrantee period in server service, and now we are getting an additional three years from them as desktops for our production staffs. Admittedly, K8N-DLs need to have their Southbridge fan replaced after three or four years, but they seem perfectly capable of working hard even after the fan has died. (I did buy a big ol' bag of fans a while back, and we replace any dead ones we discover.)
I had just enough of these older servers to build all the required production systems, with a cold spare for each province. I know some people would freak out at the concept of running out of warrantee gear...but the K8N-DLs were from the days when ASUS wasn’t a complete [long string of exceptionally vile expletives] and as such I have faith they’ll keep on ticking. I’d go so far as to say that the K8N-DLs may have been one of the very last good ASUS boards ever made. (Though the P5E-VM-DOs come damned close.)
Though my existing fleet of servers and desktops is almost exclusively ASUS everything, it’s the last generation of such. I’ve been screwed over by ASUS enough times that tiny budget or not, it’s time to start stumping up for Tyan and Supermicro. None of this “here’s a BEAUTIFUL server board. Build your fleet on it!” Followed soon thereafter by “oh, by the way, we won’t release a BIOS upgrade for the next generation processor, instead we’ll release a slightly different model name that’s IDENTICAL IN EVERY WAY, except it has a new BIOS!” (Doubly frustrating when Tyan, Supermicro and even GIGABYTE of all companies release BIOS upgrades for their serverboards of the same generation.)
Yeah, ASUS can [something suggestive that makes 4chan blush.]
Anyways, where was I? Oh yeah; old systems. Well, our full-fat clients are naturally tied to the server, not desktop refresh cycle. At the time of the last server refresh the vPro systems were just coming into view, and they cost us about $1750. I think we were able to convert the old servers for about $600. So we saved a pile by going halfers with our fleet. (To put it in perspective, the money saved this way enabled me to buy our very first shiny new UPSes. You pick and choose what to sacrifice when your budget is close to the bone…) The production staffs themselves use the old converted servers, and the “client machines,” (those systems provided for walk-in customer use and abuse) got the vPro systems. Got to make sure the customers see the best we have to offer.
So, yeah. FUN TIMES.
On the other hand, I can legitimately say that my specialities in IT are a) making computers do things they were never designed to do out of sheer necessity and b) discovering new and interesting ways to keep costs down while still providing all the requisite functionality, redundancy and reliability. Gotta be worth something, no?
- Apple: We'll unleash OS X Yosemite beta on the MASSES on 24 July
- Pics It's Google HQ - the British one: Reg man snaps covert shots INSIDE London offices
- The END of the FONDLESLAB KINGS? Apple and Samsung have reason to FEAR
- White? Male? You work in tech? Let us guess ... Twitter? We KNEW it!
- Put down that Oracle database patch: It could cost $23,000 per CPU