Hi,
I originally posted this in the Exchange forum, however as the requests are triggered by the client I have been asked to post here as well, as someone more versed in outlook may be able to explain.
Description of the environment:
multiple Exchange 2010 SP3 RU12 multi-role servers
Windows Server 2003 domain controllers
Mixture of XP clients with outlook 2003 and windows 7 with outlook 2010
Description of issue:
When users sent an email on Outlook 2003 on XP they experience short delays, perhaps around ten seconds, if sending to a large DL. Since migration to outlook 2010 on windows 7 this delay has increased to 5-10 minutes. We noticed that the issue was with NSPI
throttling, which we resolved by removing the "percenttimeinAD" throttling policy on the CAS, this decreased the delay to around 30-40 seconds.
What confuses me is that when using wireshark to monitor outlook 2010 the delay is all caused by a lengthy exchange of NSPI requests and responses, indicating the client is recursively expanding the DL when "send" is clicked. However when the DL is
expanded in advance, it expands instantly, and even the EWS requests for the mail tips responds instantly also. So what is different about clicking send that it spams the server with so much NSPI traffic? The message itself is only received by the exchange
transport with the email address of the DL? So why is it expanding the DL on the clicking of send? Hiding the DL members from the client resolves the issue, but is not practical for us to implement that.
Some outlook 2003 clients we have do not seem to exhibit this NSPI traffic and the message sends instantly so is there perhaps some Outlook 2010 GPO that we can set that prevents such an intensive directory operations when send is clicked? The delay appears
to be with the DNtoEPH NSPI request.
The performance counters show peaks in NSPI browse latency of 57 (I am assuming units is ms). The domain controllers themselves are performing normally in terms of CPU/RAM etc.
The domain controllers are due to be upgraded however there appears to be some client side behavior exacerbating this that I hope can be controlled via GPO. The client is asking lots of information about the DL recipients that it doesn't need to know, and it
is not related to mail tips as the EWS request completes very quickly, and so does standard DL expansion when clicking the + symbol.