Right lets try this again, this time on a fixed connection and not sat on a train.
I started to ask a question regarding purchasing orders for rack units when I am more of a desktop person.
I need some information regarding 1U units for price/power.
We need to do testing on our main product. Currently we use 10 desktop PC purloined over time so they are all below spec. They run 24x7 on automated test scripts. What we want to do is test feature branches before they get merged into the main source.
We need to test XP, Vista, Win7 (32 and 64 bit) running as VM's and as many as 5 branches (1 per team + 1 main tree) so that is 5 times as much power as we currently have.
We need to automate much of the testing environment. Currently we need to ghost the desktop machines and get the latest software on there and then start the automating tests.
@graham.reeds Yeah, this is all well within the remit of a VSphere setup, but you need to have a real think on how many machines you need to run at once, how RAM they'll need etc
@HaydnWVN Maybe, but this chat isn't really a place for live support. If you're having an issue, best thing to do is to ask a question on the main site. If you have already asked a question, paste a link to it and we'll have a look.
@graham.reeds Maybe work a little backwards - what's your budget. It sounds like you may be able to work with VMWare Essentials, but it'll limit you to 3 hosts, with 2 physical CPUs per host
Depends on the budget, one Dell R710 or a DL360 fully specc'd would probably be fine. Of course you might want two or three for load balancing/ fault tolerance.
If you want high end features then VSphere can quickly become the most expensive part. That said, Essentials only (The lowest one) licenses you for 3 servers, 2 CPUs per box for about £500 per year
Essentials Plus would probably be "nice" for you to have, as it comes with VMotion but from what you've said, I don't think you need the features
@graham.reeds the others speaks the truth. number of physical hosts has a strong impact on your licensing costs. Why not get faster hardware instead of wasting it all on licenses?
and yes, 50 VM's with 1vCPU and 2GB ram is -nothing-, as long as those 50 vm's aren't doing something at the same time
you'll get far with two modern specced DL360's (1u, if space is a consern) or DL380's
and honestly, all the testing you're describing isn't solved by virtualization.. you still need something to push out the correct software with the correct amount of updates
the other nice thing about vsphere is that you can set the vm disks to non-persist. It basically means that every change is reverted when you power off/on a vm.
yeah, both vsphere and hyper-v has a extensive set of scripts available
@graham.reeds Realistically though, whats is your budget? I reckon 3x high spec servers is the way to go but you can easily spend some serious money on them
@tombull89 I'm just thinking best way to do the physical disks. Most VDI deplolyments are essentially read only, save some caching so two 146 15ks in RAID1 is fine, but if it's 50 VDIs hammering the disks then it's going to grind
@graham.reeds We're not being awkward dude, but speccing for virtualisation is VERY workload dependent. It's almost impossible to give specific advice without knowing exactly what kind of work these machines will be doing
Trust me budget is not a concern. If we need $50k of hardware and we can prove we need it then it won't be a problem. If we ask for $50k and we can't prove the need then that is a problem.
The tests are running a scripted tests for a large desktop product.
What the tests do is input data and check, from a various screens that the numbers are correct.
@graham.reeds But what do these scripts do? Is there much read and write activity for the disk? Do you they do a lot of memory swapping, is 2GB going to be enough RAM or will they be swapping to disk in no time. And what about CPU useage, is the processing CPU intensive?
@graham.reeds The question here isn't what functions do the tests run, but whether the test suite overall, when running, is consuming CPU, disk, network, RAM, etc., and how much. It's possible you've never needed this data before.
@graham.reeds As very broad advice, my starting point would be something like a DL360 G7, with a pair of Intel X5650's, a 1GB write cache, an extra NIC and 4x 15k 300GB disks with as much RAM as you like. Two or three of those would see you very nice
But it's very hard to say without real knowledge of your apps. Before comitting, you probably need to do some performance monitoring on your tests and get an idea of the metrics we've mentioned
@graham.reeds Yep, a pretty good starting point. I'll hold my hands up and say it's not really something I've had to do too much, but I'm sure others here will be able to help you a bit more on that
@graham.reeds Just my brief opinion on what you've said here, there are way more knowledgeable folk, especially when it comes to HP and Virtualisation. @ewwhite and @Chopper3 to name two
You might of done. I was on a train and that disappeared out of reception and then I was home and home life rules until I was back at my desk the next day.
He said earlier that a $50k budget was "no problem" since this was a investment for many years ahead.. I see no reason to cut everything to the bone...
especially when it comes to local storage vs centralized storage, as the ROI of the local storage would basically be 0 if you move to a shared solution
@pauska Agreed, but what i find slightly odd is if they've got the time to re-Ghost those machines by hand, yet an 'unlimited budget'... Those two things (time & money) rarely come hand in hand
For example here - i have lots of the former (to a degree...) and not much of the latter... They'd rather me continue supporting windows 2000 clients than replace them.
(but i dug my heels in and got replacements anyways)
@pauska It's a fair point and yeah, I'd probably recommend essentials plus but it's probably unfair to say he needs it. But £30k doesn't get you THAT far into the virtualisation arena
@pauska: The VM aren't going to be kept. Failure of a machine is okay. We currently have a bunch of machines and regular human interaction to keep things running. We need to start running the scripts for feature branches. This instantly doubles the resources for 1 additional branch.
While a machine failure may delay results for some or all branches we will still get results. Prior to release there will be no new features going in so the changes will be minor and capacity will be spare. So failure in that scenario while not ideal is survivable.
We are looking at about 4 or 5 concurrent branches. So we need a lot of extra capacity.
Just keep in mind that if one of your PC's goes boom today, you still have other PC's to image, and you also get to keep your images (templates in virtualization)
if you have a single server local disk storage, if that server goes down you loose -everything-
@HaydnWVN: We've done it this way for so long and each time the say problem bites us. Branches weren't seriously tested when they have been merged - just general dev testing - and gotchas appeared it was a nightmare to sort. We want the branches tested as well as the main branch.
@graham.reeds ok, so this is a VDI type solution you're looking for right? your own code, what's it's OS/load-profile?, is this for production/end-user use or for testing? what level of resilience do you need
@graham.reeds VDI stands for "Virtual Desktop Infrastructure". It's generally meant to refer to using virtual machines remotely with thin clients and such, but it fits your use case as you're not really doing what people would consider to be server virtualisation
so this is JUST for testing - cool, what are your resilience requirements do you know? if you have one server and it goes pop how long could you manage with it down?
If we had 1 server and it went down just as we are doing final acceptance tests it will be panic stations but to get up to 30 VMs running that looks to be several machines which will give us spare capacity.
There is one exception that we do a torture test with a very large dataset for timing purposes, but I guess that will still be run on minimum hardware.
@graham.reeds ok, well load-wise this isn't much to worry about bar possibly a boot-storm on the disks, do you want a pizza-box server or do you have any other physical server requirements (i.e. blades etc.) and do you have a budget already
@ewwhite That is terrible yes :) I have an 8 host cluster at home
@graham.reeds so you want say 2 x vCPU at 100% for 30 VMs? fair enough, I get that, while I normally would recommend Intel Xeon based servers perhaps just two of HP's DL385p Gen8's would be very cost-effective. Put two x 6278 Opterons in them, 64GB would do easily and then stick in the second disk bay and fill it with a bunch (12-16) of their 146 or 300GB 15krpm 2.5" SAS disks in R10 with vS 5.1 on it
@ewwhite the web-based client stuff is...different
was playing with the replication stuff but ran out of time
Im in DMZ mode so Im firewalling myself, stealthy all ok, but
get faulty test results from shields up that there are pings...
yesterday I couldn't make connection to gameservers work, because ping block was enabled(on the router).I dissabled it, but this persist even due my firewall.
what is the ...
I'm looking for something special : here is a windows server that is used for development by 4/5 people remotely working on it through remote control.
I'm installing Visual Studio 2010 on it, and the installation won't continue until the server has been restarted. I'm looking for a way to reboot...
I'm sure he could script what he wants with minimum fuss, but it's a shonky thing to just reboot a server anyway. What if someone just disconnects for a second, or they're about to jump on and do some work while it's going down
@MichaelHampton It is a crap question, though. Even with 50, 60, 70 users logged on it's still far better to arrange a window with people than it is to just bounce it as soon as no one is looking.
@ewwhite yep, eight 460 hosts, a pair of 460's as a W2K8R2/MSSQL cluster, another pair as an Oracle RAC cluster, and a BL2x220 pair for admin/util stuff (InForm etc.)
Just got in a server order configured by a sales guy that no longer works here. Wrong hard drives, wrong redundant PSU (the model he ordered can't even do redundant), and the parts are all opened.
@Chloe Hahaha... "I'm running Windows 2008 DataCenter Edition… and I need dyndns." I suspect he didn't pay for it.
@ChrisS Yes we do. Started doing so after the first time this happened and we ended up with $20k in hard drives that are now in various home machines. The tech guy who vetted it slipped up. But I like blaming the sales guy :p
I have some newb questions I want answering please about cloud hosting - we are currently looking at Rackspace and getting a windows box. This is the situation:
We have 15 computers in our office.
We have 3 printers, some wifi and some network plugged.
We have a standard router and the office sh...
That there is fucking hilarious and everything that is wrong with using the phrase "cloud"
"Can we have a server in the cloud that does this? Is it possible I mean the computers would be wireless connected to this cloud and you turn the machine on and its hosted?"
The real irony is that he says "We have a network company saying..." like their full of shit. They've probably specced him up a nice small office system, but it's just not cloud enough
I wonder how much it'd cost to build an orbital data centre connected to earth by a nano-tube based tether, cooling would be cheap, it could be solar powered and more importantly it'd give the vacuous senior managers who only learn about new things from their inflight-magazines something new to bang on about as it'd be "above the clouds"
@Chopper3 Cooling in outer-space is very difficult, as you can only use radiation, convection and conduction don't work (which are the two ways we normally use on Earth).
I have been tasked with finding/creating a bulk fax solution to fax individual reports to hundred of our clients (each report will be unique). We would prefer to stay away from a per page service as the cost would add up rather quickly.
The key issue is we need to send all of these faxes sent a...
If you tossed a human out into space, and assuming they were trying to survive, they'd suffocate before dying of anything else. If brought back "inside" in less than ~2 minutes they'd heal up, no permanent problems. Staying "out", they'd take over a day to freeze solid, though the moist areas of the body would freeze within several seconds.
@Chloe Large pressure differences can do really spectacular damage... Most people assume space is a significantly different pressure than air-pressure down here.
@Chloe For instance, when working on a trans-Atlantic (might be Pacific, not sure) pipe, a thousand feet (300+m) down (water pressure increases ~.445 PSI/ft or 3kN/m²) the pressure is enough to do extraordinary damage youtu.be/LF-lFd-NUI8
@Chloe It really doesn't matter which way you're going... The pressure in that crab's body was the same as the water around him. Inside the pipe was less. Similar to a human in space, but on a different order of magnitude.
@ChrisS Actually, would a person who got tossed into space freeze that quickly? The specific heat of a vacuum is pretty low; I'd think it would take quite a while to radiate off all the body's heat into a "medium" that's essentially a perfect heat insulator.
@HopelessN00b It does depend on a lot of things, but the human body has enough water in it that the water would steal a lot of heat freezing and sublimating. It take a month before the body got to any significantly negative temperatures; but freezing wouldn't take too long.
The heat from the sun hitting a body in space would actually be enough to keep that side warmer than normal body temperature (ever laid out on a beach near the equator too long; sun stroke...). So we're going to have to shade the body, or toss it farther away from the sun if we want it to freeze solid.
@ChrisS Still, without a good medium of heat exchange (loss), I have trouble believing you could cool a substance with such a high specific heat by 30 degrees in a matter of seconds, especially while it's still being heated by your core body temp.
Even hypothermia takes longer to set in when you dump someone into sub-zero temp water, and that's a much smaller decrease in body temperature.
I have been tasked with finding/creating a bulk fax solution to fax individual reports to hundred of our clients (each report will be unique). We would prefer to stay away from a per page service as the cost would add up rather quickly.
The key issue is we need to send all of these faxes sent a...
