Need to create performance baseline to troubleshoot RDC performance issues

i've been tasked benchmarking new clients current server/network configuration identify , recommend ways resolve poor wts performance.  i've not had baseline/benchmark ts environment before, , 1 has virtual ts's (extra varables).  also, information available not think i'm going have black-box project (bench , see looks out of place).

setup:

-5 physical servers running server 2008 hyper-v (64-bit)

-each server hosts 3 virtual servers running windows 2000 sp4 wts configured

-150-200 remote users (more exact numbers being calculated: users/virtsrv, users/physsrv)

-data center internet connection 5 mbps (up , down)

-remote site connections typical dsl (5 down, 0.5 up)

the information releasing still vague, please bear me (feel free recommend questions should ask).

users intermittently (but often) experiencing performance issue when running programs in ts session.  exact descriptions still vague (ie: "it's slow").  no reports yet of disconnects/timeouts or 'soft crashes' (program crashes, timeouts, etc), general poor user experience.  issue seems affect given day of week, tue-thu noteable (probably users don't 'care' on either end of week).  far issue seems come , go, cannot reliably walk in on given day/time , witness issue.

i suspect biggest performance issues related amount of upload bandwidth have @ data canter (5mbps).  spread on 100 users 50 kbps/each.  try gather bandwidth stats ips/firewalls see if can rule out.

i don't know if when network , wts servers installed if benchmarked or not (server cpu, i/o or ram loads), have treated "unknown" now.

i'm trying narrow down if run big daily reports or not (ie: report takes 2 hours run, user triggered , no specific start time), these might saturate cpu or i/o of vm's , hosts.

one other possibility might when have shift change, , previous shift doesn't log out, might causing ram overload on vm's (memory swapping) until idle or disconnected sessions expire , terminate.  comments on if worth exploring (or how track this)?

i did googling perfmon stats track i'm not best locations track crux of why i'm posting here.

i think need place disk queue and cpu load tracking counters on servers (physical , virtual) , network throughput tracking on physical servers.  feedback on if choice start?

i'm looking for some means gather 2-3 days worth of stats put in chart (or something) , see if patterns or bottlenecks can identified, i'm open additional recommendations too.

although article there while it's still valid start troubelshooting performance issues on ts/rds environements. hopes points in right direction http://sbc.vanbragt.net/mambo/index.php?option=com_content&task=view&id=42&itemid=192.

---

http://sbc.vanbragt.net

Windows Server  >  Remote Desktop Services (Terminal Services)