-10

Desription of problem changed : how to make sure, that my Hyper-V is using all my hardware resources, and if it is using, what is the bottleneck of the system, when discovering performance issues? - First of all what is the rooftop of the hardware performance (so I could compare physical machine vs hyper-V Virtual machine) - Then how to find what is causing my Server slow performance - how to speed up the system?...Hyper-V has ability to use only 4cores, too bad even for 2x quadcore CPUs, not talking about cluster...I found possibility to make it force to use whatever cores I need, but : I did not discover any significant speed change, is it still limited to 4 cores..? I found many Things-to -disable as NUMA nodes, VMQ, dynamic memory alocation, etc.....which should help, I guess there must be much more, if my server is getting bored(CPU,RAM,HDD), but users discovering great latency. Any idea how to measure what is real (in Hyper-V or elsewhere)? Thanks for any ideas....

Jozef
  • 61
  • 1
  • 2
  • 7
  • MSSQL doesn't play nice with other services on a box. It should be someplace else....that's just touching on the issues here. – Nathan C Oct 20 '14 at 16:27
  • 3
    I am sorry, but your question is a rambling TL;DR mess. Can you edit it to focusing on a specific issue? – Zoredache Oct 20 '14 at 16:30
  • 5
    So, you're engineering a huge single point of failure for your entire workforce? Moving to an RDS based environment is fine in principle, but requires a lot of care in practice - having MSSQL running on the same box is no doubt the first of many big mistakes here. If you're doing some kind of development, VDI is probably a better solution anyway but is more costly. – Dan Oct 20 '14 at 16:31
  • 10
    There is so much wrong here... I don't... I can't even... I mean... you know what, I'm just going to chose to believe this is a brilliant troll. Kudos, sir, on one hilarious post. – HopelessN00b Oct 20 '14 at 16:32
  • 2
    @HopelessN00b - Off topic and no offense to the OP, but your comment had me in stiches. :) – joeqwerty Oct 20 '14 at 16:34
  • Hi Guys, thanks for critics BUT:as I wrote before:This"mess up" is Stage 2-all merged on single hardware,to test performance of Hyper-V+several users.To reduce influence from other things(switch,ethernet traffic-RDP load,firewall settings,This PC,that PC,driver compatibility,etc.)I just wanted to reduce the no.of Bottlenecks and waiting times for external"UNKNOWN things".MS SQL and other things should be on dedicated machine-I know that,as nobody could give me Answer:how super-PC schould be build(single OS on scalable hardware-Blade,Hyper-V,or..?)I made these steps to find out the bottlenecks – Jozef Oct 21 '14 at 19:20

1 Answers1

4

I'll bet any amount of money that one of the problem described in your second paragraph is your disk IO. Check your Disk Queue length in Perfmon on the actual hardware server. If you're using a RAID controller without an enabled write cache (Usually via a BBWC or FBWC, battery or flash-backed write cache), you're doing RDP wrong.

Why do you also have MS SQL on a Terminal Server for interactive users? Get that onto another machine.

mfinni
  • 36,144
  • 4
  • 53
  • 86
  • 5
    In addition, I suspect his disks are SATA which is also a bad recipe. For any decent amount of density you should be looking at 10k SAS at a minimum. – Dan Oct 20 '14 at 16:40
  • Hi, Thanks for productive answer. The RAID10 is 4x identical SAS 15Krpm,+ SSD for holding OS and Database...IOps for HDDs seems fine, but I will check that tomorrow when full load will. The thing is : - there is not a tool to simulate a realistic 100 or 200 users on single PC. - I could not find any straight answer, how super-PC should be build - RDP takes me 20Kb/s/user, so here my 1Gb and later 10Gb LAN will be OK for while - so RDP way of work seems good solution - I need single scalable hardware, ideally blade – Jozef Oct 21 '14 at 19:27
  • Anyway: I read some good posts about the PCI SSDs, but to be honest : The reply from SQL server is not slow, the HDD is normally not slow (I can copy 10Gb files on backround almost normally) The slowishness seems to be related to NUMA nodes, OS, Hyper-V, or some Software limits - I have slow even Chrome tabs, not only SQL, or HDD related tasks...Is there any topology of : how the hardware/software should be set and used for load of 20users, which will be scaled later on to 100 or more users?..I am talking still about the RDP or PCoIP, etc....single Hardware for all users....any idea? – Jozef Oct 21 '14 at 19:33
  • I didn't even whisper the word IOPS - I said "disk queue length", which is quite a bit different. – mfinni Oct 21 '14 at 20:42
  • Stop using the phrase "Super PC." Out here in the professional world, we (and microsoft) have been calling it Terminal Server for about 14 years or more. – mfinni Oct 21 '14 at 20:43
  • HI, Sorry, for Disk quee : bellow 0,05, IOPs which I added too (as I read that this might be the main problem) is well in HDD limits too(sorry for adding this information) SUPER-PC thing : I dont call terminal server a SUPER-PC, a SUPER-PC from my point of view should : 1, connect 100+ users through RDP 2, allow full HD video with audio playback 3, allow a non-standard programms run for most of them (CAD/CAM, 3D modeling) – Jozef Oct 22 '14 at 14:26
  • 4,have sufficient CPU power to calculate some CPU intensive simulations ..1PC:with as much RAM as needed for everyone,with CPU power of graphical station for every user granted – Jozef Oct 22 '14 at 14:33
  • You can make up whatever terms you want; don't expect the rest of the IT world to accommodate you. – mfinni Oct 22 '14 at 14:53
  • I appologize for bothering you, if so. As I have no idea if "known phrases" as RDP, Terminal server, Hyper-V,etc. are suitable(I never build a very powerfull centrall piece of hardware)I was only looking for somebody to help me : – Jozef Oct 22 '14 at 15:32
  • Understand: - confirm if my hardware will have always that big "performance offset" due to Hyper-V (possibly find some steps to allow better use of hardware) - exchange some advices on what to do/not to do, when building one single powerfull OS on scalable hardware, instead of many smaller VMs. Maybe I placed the question not the right way(I am not that good in Windows servers)I would like to change it,and start normall conversation, but I dont want to bother anybody. – Jozef Oct 22 '14 at 15:33