cluberti

Yes - Exchange 2000 needs to be installed on a Windows 200x server in an Active Directory domain. You can run Exchange 200x in a mixed-mode AD forest with a directory connector to Windows NT 4.0 domains, but this is definitely sub-optimal and should be avoided when possible.

It would be interesting if you had a network trace from the client of the explorer slowness - since this doesn't happen in a command prompt, it's likely extra commands sent across the wire by explorer.exe to do something, like parse directories, display certain types of folder icons, etc. You aren't using Access-Based Enumeration (ABE) on that 2003 server, are you?

How do you do Windows Update control? My domains all use WSUS 3, and no one got the update because I've configured it to not do anything with new updates delivered during sync (they were all "Not approved" and thus no one got the packages). Those who are seeing this either had 2.5 or 3.0 already approved, or had the "approve optional updates" selection turned on (instead of set to just auto-approve critical updates, or have it configured not to auto-approve at all, which is the best setting IMHO). I haven't checked my home machines though, so they may have gotten it because I wasn't at the console to say "no". With 350 machines, I would strongly suggest using a WSUS server to control updates if you aren't already.

And you are correct - my math skills suck again!!!

Having a paging file at the beginning of the disk (the outer sectors of the platter) is actually faster (not by much, mind you, but it is faster). Here's an older article talking about why sectors on the edge of the disk platter can be accessed faster than inner sectors of the platter, although the differences are likely not as great now (speed-wise) as they were a few years ago, when the article was written. Having files stored in sectors at the beginning of a disk (at the edge of the platter) does indeed allow faster disk head access to the sectors holding those files than would be if they were in the middle or end of a platter, but again, the performance difference is likely very small unless you have a slower laptop drive (7200, or especially 10K RPM drives, probably wouldn't show a noticeable difference in day-to-day speed to the page file no matter where it was on disk). It's important to note that one of the main reasons that the built-in defrag engine in Windows doesn't move the page file to the beginning of a volume is due to the MFT slack space reserved for MFT growth, which is located at the beginning of a volume (it's a set size, based on volume size) and no one is "allowed" to overwrite this space unless there's nowhere else on disk to write a file when a write occurs - in which case the MFT can give up some of it's slack space for the write. Also, since the default setting for the page file is dynamic, the OS doesn't assume that you'll specify a max size for the paging file and move the MFT slack space, and thus it does assume your page file may grow at some point and thus it cannot be before the MFT or MFT slack space in this scenario (because this should never move, technically) without the guarantee that the page file will defrag, which Windows does try to avoid (although poorly). Also note that while having the page file at the beginning of a volume will speed up access to those sectors, it won't make your seek time faster, so actually reading and writing to those sectors isn't going to be faster, just getting the head there when your disk does find that the page file is there will be (and coming back to files the OS uses more, like the MFT and system files, will be "slower" because THEY aren't near the beginning of the disk ). All in all, the most important part of this is that the page file be defragmented and in one contiguous file to increase speed and reduce seeks for sections of the page file in different physical disk locations, but the actual physical location of the page file on disk really shouldn't matter as far as perceptible speed unless you have slower disk drives. I'll state it again - location of the paging file (especially if you have a lot of RAM and a small page file to limit the possibility of a commit of VA to RAM) really shouldn't matter much if you have faster IDE disk drives, but if you really want to move the page file to the beginning of a drive, though, I know of no free utilities to do this. Diskeeper (pro versions), UltimateDefrag, Norton SpeedDisk, and possibly O&O defrag can do this, but none are free.

It'll tell us what happened when the process crashed - once you are able to get a dump of the crash, upload it and post links so we can grab it and debug it for you, if you'd like. At this point, it'll be hard to say otherwise why this is occurring other than an educated guess saying the intel driver is at fault somehow.

Do you see the same sorts of delays when browsing the shares via a command prompt? Also, are these mapped drives (drive letters) or just browsing to \\server\share?

You can get to the admin password with ntpasswd. You can click the "bootdisk" button, and then download a floppy image which can be written to a floppy, or a boot CD image that you can burn to a CD. It's not the prettiest out there, but it's (at least IMHO) the best and most reliable (aka won't hose your system).

Where is the folder on the Vista system - is it in the root of a drive, or somewhere other than the "Documents" folder in your profile?

You need to run perfmon and add Memory / Commit Limit and Memory / Committed Bytes to see these data points in Vista. They are no longer in task manager.

If you can, get the PID of the svchost containing that service, and then install the debugging tools and get a dump of that svchost process ID when it crashes.

Technically, DFS should be able to do this for you if both servers are on the same domain structure - the article linked below is specific to RIS, but it can be generalized for any and all DFS replication schemes. I'm not sure what you mean by DFS not meeting your needs, as what you specify (folder replication between hosts) is exacly what DFS is for. Perhaps I'm missing some tidbit of info about what you're replicating (I'm assuming you aren't going to be using DFS to replicate the SQL database, but only files in folders): http://support.microsoft.com/kb/273594

I track load times on apps and file system reads/writes via perfmon, and file copies with a vbscript - I've noticed that a machine with 1GB of RAM or less (especially 512MB) seems to perform noticeably better in these tests than boxes with 2GB or more (1.5GB seems to depend a lot on hardware - some machines benefit, and some do not).

I really like geek's sense of humor - good to have you 'round, mate .

Personally I have to disagree here - Windows should NEVER just BSOD, ever. If it does, don't ignore it .

Have a read - I've found that on systems with 1GB or memory or less, USB and fast SD cards do make a difference, but when my boxes have 2GB or more it seems to matter very little that readyboost is enabled.

/me chuckles

If you created a new domain controller to replace the old one, but didn't specify that the second DC (the new server) was a GC server in AD sites and services, and you also didn't allow time to replicate, the old GC data is gone unless you have a system state backup from the old DC you can restore (from before you demoted it). You can indeed specify the new DC is a GC in it's object's properties in AD sites and services, but you will have lost old GC data - it is gone.

Yes, I'd like to see the dump itself - I can see some netbt traffic there as well, so I'm curious. I'd like the opportunity to interrogate some of those addresses in the dump file itself.

That was the stated reason, that they were skimming a profit above and beyond media and shipping costs. I don't know if it was true or not, because obviously I don't know how much it cost them to produce these, but that was the reason stated. I'm pretty sure that other utilities that point users to the microsoft download servers don't have to worry about being "gone after", but projects with the visibility of AP and supposedly making money off of Microsoft code (patches), that's gonna get you shut down.

If you can press CTRL+ALT+DEL and open task manager, is explorer.exe listed as running? If not, try launching it with File > New task...

No, 4GB is the limit for x86 because 2 to the 32nd power is 4GB, and that's thus the max addressible range for RAM on x86 (x86 processors us a two-level page translation scheme, thus the 2). Intel introduced the PAE extensions to the PPro line of CPUs to bridge the gap between x86 and 64bit, but it took a bit longer than they had initially hoped to get to 64bit computing . One fun sidenote on that is that the 400MHz bus Pentium M processors do NOT have PAE support, which is very odd and interesting.

I could guess, but I'd rather see a dump file to analyze. Can you configure your box for a complete dump and provide that the next time it occurs?

Did you use ICS or RRAS to set this up on the DC?

Are you sure that in your DNS console that you actually have a RECORD in DNS that is an SOA record, containing the data you entered in the GUI? You can put the data in the GUI all day long, and if the SOA record is incorrect or missing you will get this error.