|
Speaking of which Steve........
I am starting to see questions of the type of how does 64
bit DC change the best practice 4:1 proc recommendations for Exchange to GC
processor. Does PSS/MCS/Dev have any thoughts? Especially if you are able
to cache the entire DIT. I have seen some 64 bit testing numbers from third
parties but that is far from authoritative in terms of what MS thinks for the
best practice numbers which weigh heavily with customers who want to do it the
"Microsoft way".
Ditto the dual core CPUs.
Another one that recently came across my desk was if you
have 4000 users on a 4 proc Exchange server and are currently using a single 1
proc GC and then you decide due to load on Exchange (say RPC load due to
search/archive software which isn't impacting GCs) you want to go to 2 4 proc
Exchange servers with 2000 users each do you have to go to a dual proc
GC or add another single proc GC or is it ok to stay with the one single proc
GC?
Oh and another question I was asked was about using single
proc GCs versus MP GCs and how the scaling of MP wasn't linear so should that be
somehow involved in the Exchange best practice numbers?
It seems from my experience that you do better with making
bigger and more powerful GCs in general because while Exchange does
some limited logic round-robin load balancing at the server level, it doesn't do
it at the site level amongst all Exchange servers so you can really start
beating down a few GCs while the others see relatively light loading. Of
course you don't want to have few GCs though in case you do have a problem so
you throw a couple of extra larger GCs into the mix for fault tolerance for when
you have to bring a GC down for maint or it just falls down for some reason.
Also it seems that there is no real good way of determing
exactly when you need to change your GC strategy for Exchange because your
various Exchange AD related counters could be poor yet AD is still seeming to be
performant and possibly even under utilized. This seems to really come into play
if a lot of DL expansion of very large groups is coming into play. Possibly it
is simply related to bad queries from Exchange due to, well bad queries, or
third party event sinks a la Exclaimer or multiple to software, etc.
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of Steve Linehan Sent: Friday, October 14, 2005 1:25 AM To: [email protected] Subject: RE: [ActiveDir] Domain Controller Consolidation utilizing Dual Core CPUs In my opinion the
biggest bang for the buck is consolidation of servers to the 64bit platform
assuming of course that you have a large enough database, greater than 3 GB, and
put enough memory in the servers to cache the entire database contents. I
have come across very few cases where Domain Controllers were truly CPU bound
and in almost all cases they were I/O bound. These servers perform
extremely well for servers that are taking large amounts of ldap traffic from
applications like Exchange. Thanks, -Steve From:
[EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
On Behalf Of Mauricio F.
Funes Gentleman, Thanks for your input,
Mauricio
Funes |
Title: Domain Controller Consolidation utilizing Dual Core CPUs
- RE: [ActiveDir] Domain Controller Consolidation utiliz... joe
- RE: [ActiveDir] Domain Controller Consolidation u... Rick Kingslan
- RE: [ActiveDir] Domain Controller Consolidation u... Ken Cornetet
- RE: [ActiveDir] Domain Controller Consolidation u... Thommes, Michael M.
- RE: [ActiveDir] Domain Controller Consolidation u... Roger Seielstad
