All of this brings you back to the *root* of the original question. In what situations does it make sense to increase the replication frequency?
And I think we have an answer here: First we acknowledge, that a low frequency does not reduce the number of changes that need to be replicated, unless there are short-term temporary changes that can be eliminated from replication (such as group created and then quickly deleted before it ever replicated.) Second we acknowledge that the total traffic for replication is linearly determined by the number and type of changes, plus some overhead for establishing and tearing down the connection. So a higher frequency replication produces more bandwidth consumption in precisely three ways: Increased overhead for establishing and tearing down connections, decreased efficiency of compression when a smaller number of objects are being replicated, and increased payload for situations where temporary objects didn't need to be replicated. The question is, how can we determine the significance of these three increases? There are two ways I've been able to find, to measure bandwidth consumed by replication. Either go to perfmon, and watch the NTDS / DRA counters, or build an isolated DC whose sole function is to enable you to measure this. If you give a DC at some remote site only the role of DC, and you use wireshark to capture traffic to/from other DC's, that should be an easy and reliable way to actually measure the replication traffic. I'm in progress right now, building a DC at a remote site. I should be able to measure the bandwidth with that machine soon, to serve as a cross-reference, sanity check of the NTDS DRA counters. Measurement #1: Right now, I've been able to measure 7K bytes used to replicate when no changes occurred, using perfmon. And I was able to measure 11K used to replicate a new user account creation or deletion. This is on-par with the object sizes that are listed in http://technet.microsoft.com/en-us/library/bb742457.aspx#ECAA Pending the acquisition of more information, I'm going to have to say, 7K every 15 minutes is nothing. So from this standpoint, the information supports pretty much always increasing the frequency to max. Measurement #2: I'm going to go out on a limb and conjecture that compressing one or multiple objects does not significantly affect the overall compression ratio. I'll be looking to test this conjecture by measurement, but I'm pretty confident right now declaring the difference is not significant, and again conclude that increasing replication frequency is generally a good idea. Measurement #3: The number of temporary objects that otherwise wouldn't need to be replicated is organization-specific, determined by the behavior of your sysadmins. So I can't make any generalization or conclusion on this subject. I can say that for the organizations where I've worked so far, this traffic would be negligible, but it could be significant for other organizations. _______________________________________________ Tech mailing list Tech@lists.lopsa.org https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
