parallel scaling is hard. without a load balancer there's no really trivial way to scale out. upside is a load balancer wouldn't be that hard to do (in linux ;-> but then that machine is beefy enough you could give a linux VM 512mb of ram and have it load balance and cache stuff without an issue) some kind of multi-master sql replication to keep the database content in synch between the two machines would work well too. For webfarm stuff the general idea is redundancy at the application level, virtualisation is a way of tacking it onto stuff that's bad at it on its own. I'd perhaps look at running the webserver on the old machine and the database on the new machine, see what kind of processing loads your actually using to see whats worth doing. SBS only lets you run one instance per machine (as i understand it, I don't have my PHD in microsoft licensing though) So you would need a licence for each physical machine (in theory), though precisely what they would do about it is another question. The second machine would be a virtual machine host (virtual server has different connotations) as would the primary machine. hit me up by personal email if you want a more timely response lol. On 03/06/2012 02:46 PM, James Newton wrote: > I guess I really should talk to you about this virtualization thing... > > Is there a way of setting it up on one machine to start, and then expandi= ng > to a second machine later? The thing that kills me about this is that as > soon as I get the new server running, I can format the hard drives on the > old server and it will probably be good as new... I could sell it at a lo= ss, > or keep it as a backup... OR.... I could make it a second virtual server? > And run my one SBS 2k8 license over both machines? Or am I not > understanding? > > -- > James Newton > 1-970-462-7764 > > -----Original Message----- > From: piclist-bounces@mit.edu [mailto:piclist-bounces@mit.edu] On Behalf = Of > Jake Anderson > Sent: Tuesday, February 28, 2012 18:48 > To: Microcontroller discussion list - Public. > Subject: Re: [OT] New server comming, Request For Comment > > rofl > that'll learn me for not paying attention to the piclist > feel free to drop me a direct line about this kind of thing. > > i stumbled across this too > http://lists.us.dell.com/pipermail/linux-poweredge/2010-May/042140.html > which might fix your "punctured stripes" > > btw that raid is fakeraid as far as i'm aware, i wouldn't use it, has > the downsides of hardware and software raid combined. > I *think* win2k8 can boot from software raid1, I'd do that, even if it > cant, put your data on software raid, the added CPU load should be > pretty much nothing and you will have cpu to spare, thats a beast of a > machine ;->. > at least with software raid you can stick the drives into another > machine and suck the data out of them. > > I'd also look at putting your iis and sql servers into VM's even if you > host them inside hyperV or whatever microsoft is calling it these days. > see what the performance hit is like, if its not too bad then running > the site on another machine if this one dies is a matter of stuffing the > VM files into a new machine and your back up and running (or just boot > from a backup if you can handle loosing a day/whatever of data) > > I'm really liking virtualisation if you havent guessed ;-> it just makes > everything so much easier in terms of managment. > Its all well and good saying backup your bare metal machines, but if you > cant get an identical bare metal machine your hosed, even if you do get > an identical one mac addresses change etc and most backups don't give > you a bootable image as an end result. > VM's just make all that go away, all you messy configuration is sitting > in one file that you can drop on anything with the herbs to run it. > > On 29/02/2012 12:42 PM, James Newton wrote: >> Thanks for the feedback Jake. That sounds very nice... >> >> I've often thought of just hosting with you, but in this case, you > responded >> 2 hours after I pulled the trigger and ordered the ASUS server below. >> >> Oh well. >> >> -- >> James Newton >> 1-970-462-7764 >> >> -----Original Message----- >> From: piclist-bounces@mit.edu [mailto:piclist-bounces@mit.edu] On Behalf > Of >> Jake Anderson >> Sent: Tuesday, February 28, 2012 17:02 >> To: Microcontroller discussion list - Public. >> Subject: Re: [OT] New server comming, Request For Comment >> >> I know I mention linux here but its not internet facing so I hope it >> meets your requirements its also something of a mind dump. >> >> Any single computer system will fail in the case of that computer >> failing which is still depressingly common, thats below the SLA my >> clients need. >> >> My current "mission critical" system base is a pair of dell R210 IIs >> > http://configure.us.dell.com/dellstore/config.aspx?oc=3Dbect122d&c=3Dus&l= =3Den&s=3Db >> sd&cs=3D04&model_id=3Dpoweredge-r210-2 >> >> looks like about $1200 gets you quad xeon xeon, 8gb ram and 2x 500gb >> hdds (i'd look at getting the system without the hdd's, dell charge >> obscene amounts for them). >> >> I run linux (ubuntu) as the virtual machine host using ganeti for the >> hypervisor managment and KVM for the hypervisor itself. >> It takes care of mirroring the disks over the network with DRBD so even >> a physical machine dying isn't a big deal. >> live migrating all the virtual machines from one host to another is a 1 >> liner if i want to say upgrade the ram/os/whatever on one of the hosts. >> similar case when recovering from a dead machine. >> (there is also a web interface to ganeti, but i havent used it much) >> I don't raid the disks inside the machines i get better seek times this >> way (i have 2x disk io intensive VM's and i run one on each disk, but i >> run them on different hosts), policy is to replace the disks yearly as a >> preventative measure. >> >> live migration is nice, I have a pbx running in a VM, i've moved the PBX >> to the other server whilst it was running about 4 calls through it and >> nobody noticed. >> >> hardware raid is bad in my eyes, if stuff goes pearshaped you are >> dependant on getting that same raid controller in order to get your data >> back. >> >> This setup is nice, I have all the features you get with a high price >> VMware setup with an expensive SAN for ~$2000 in hardware, they also >> come with a BMC that lets you get a serial console over lan to them. >> >> >> For you I'd stick your iis into one VM and run sqlserver in another, >> give each of them 3.5Gb ram (or whatever seems good) and have them >> running by default on different physical hosts. >> In case of dell being dell live migrate if you can to the working >> machine, if not then fail the first machine and boot everything again on >> the second. >> >> Get the first machine fixed then migrate stuff back onto it, business >> interruption can be 0 in a good case, and a bad case is as long as it >> takes you to diagnose a machine is dead and then get it started on the >> other one. >> I like also that its not like having a "spare" machine, in that most of >> the time its hardware is wasted, this way your getting the performance >> of 2 machines when both are working, and then you just get degraded when >> one breaks. >> >> If hdd failures are your issue, then set up raid1 under it, all your >> data is then on 4 hdd's spread over 2 computers it should be safe from >> most physical failures ;-> >> >> >> >> A general observation, most of your page serving time is going to be >> seek time for your drives, have you looked at SSDs? >> They are soooo much faster for most stuff, get the intel ones though for >> line of business and make sure to backup. >> >> On 18/02/2012 8:52 AM, James Newton wrote: >>> The server is dying, long live the server! >>> >>> SUMMARY: I'm looking for review, and advice from those with M$ server >>> experience, and am willing to pay a fair amount (e.g. $100) for qualifi= ed >>> professionals who are willing to spend a few minutes looking this over > and >>> offering suggestions. This is (another) big investment, and we just can= 't >>> afford to screw it up. >>> >>> BACKGROUND/SERVER RAID FAIL: So we appear to have a busted server due t= o > a >>> "punctured stripe" issue on our raid array. Nate said this is Dell for > "we >>> sold you a cheap-assed RAID controller with awful firmware" and I'm >> inclined >>> to agree. Google "RAID punctured stripe" if you care to know more. If y= ou >>> know how to fix that without downing the server and reformatting / >>> reinstalling, we would pay good money for a solution. >>> >>> SERVER REPLACEMENT: According to Dell, fixing it would mean downing the >>> server for many days (at the speed I could rebuild it) so we have no >> choice >>> but to purchase a new server. Obviously, it ain't gonna be Dell! We've > had >>> good luck with ASUS, HP, and IBM machines in the past, so those are my >>> current preferences, but I'm open to other suggestions. I hear HP >> overheats, >>> and IBM is way over priced. We love our ASUS laptops and workstations. = No >>> problems. >>> >>> M$ ONLY: The machine must run SBS 2k8 pro with Exchange 7 and IIS 7 as >> that >>> is what we have licensed and what I know. Just to be clear: NO! *nix is >> NOT >>> an option. Period. Thank you, but no. >>> >>> Here is the current setup I'm looking at: >>> >>> ASUS TS500-E6/PS4 Pedestal Server Barebone >>> http://www.newegg.com/Product/Product.aspx?item=3DN82E16816110045 >>> >>> 2x Intel Xeon E5520 2.26GHz LGA 1366 80W Quad-Core Server Processor >>> http://www.newegg.com/Product/Product.aspx?item=3DN82E16819117185 >>> >>> Patriot Signature 12GB (3 x 4GB) 240-Pin DDR3 SDRAM Server Memory Model >>> PS312G13ER3K-E >>> http://www.newegg.com/Product/Product.aspx?item=3DN82E16820220423 >>> >>> 2x Western Digital VelociRaptor 300GB 3.5" SATA 6.0Gb/s Internal Hard >> Drive >>> -Bare Drive >>> http://www.newegg.com/Product/Product.aspx?item=3DN82E16822136929 >>> >>> Total price just over $2200. ...which is pretty good... but still > painful. >>> ,o) >>> >>> Concerns: >>> >>> -- RAID: Concern has been expressed about the RAID system. The SATA >>> Controller is an IntelR ICH10R with 6 SATA2 300MB/s ports and IntelR >> Matrix >>> Storage support for software RAID 0, 1, 10& 5. I would be running RA= ID > 1. >>> The current server with less processing power never exceeds 20% and >> averages >>> under 10. Any advice? >>> >>> -- RAM: Concern has also been expressed about the memory. The server sa= ys >> it >>> supports: >>> Total Slots: 6 (3-channel per CPU, 3 DIMM per CPU) >>> Capacity: Maximum up to 48GB (RDIMM); Maximum up to 24GB (UDIMM) >>> Memory Type: DDR3 1066/1333 Reg DIMM/ Unbuffered DIMM with ECC >>> Memory Size: 1GB 2GB 4GB 8GB (RDIMM); 1GB 2GB 4GB (UDIMM) >>> >>> The memory is: 240-Pin Dual in-line memory module 2 Rank Double-sided >> module >>> 7.8US Refresh Interval (8192 CYCLES/64MS) Auto and self refresh > capability >>> Serial Presence-Detect (SPD) PCB height: 1181(mil) RoHS Compliant JEDEC >>> Compliant Compatible Server board: Asus Z8NA-D6C Tyan S7025AGM2NR Intel >>> S5520HC Tyan S7002G2NR-LE Intel S3420GPLC. >>> >>> Customers have reported success with several ASUS systems including the >>> TS500 and the RS700. If you can see a problem, please speak up. >>> >>> -- DRIVES: Concern has been expressed about the Western Digital drives. >>> Apparently they have a reputation for running hot? This model appears t= o >> be >>> made to address those problems: "The 2.5-inch WD VelociRaptor is enclos= ed >> in >>> a backplane-ready 3.5-inch enterprise-class mounting frame with a > built-in >>> heat sink that keeps this powerful little drive extra cool when install= ed >> in >>> high-performance desktop chassis." It's also the only drive I could fin= d >>> with a 5 year warranty... But I'm very open to switching to a different >>> drive, but I want an ultra fast drive as the HD's are almost always the >>> major bottleneck in any system, and more so in a server. >>> >>> ACTION: Your input is valued and appreciated. Anyone with any comment > will >>> be welcomed. Please pass this on if you know someone who has experience >> with >>> this stuff. If you are a server professional, or know one, who would li= ke >> to >>> get paid to advise us, please send a link to your online CV or Resume. >>> >>> >>> -- >>> James Newton >>> jamesnewton@piclist.com >>> 1-970-462-7764 >>> >>> -- http://www.piclist.com PIC/SX FAQ & list archive View/change your membership options at http://mailman.mit.edu/mailman/listinfo/piclist .