Question : Dropped Network Connections

I've stumbled into this problem numerous times on many networks over the last 6 years both with Windows NT and Lan Manager. I've troubleshot myself blue in the face on and off using Technet and other sources of info to no avail. Here's the problem:

When network activity reaches a peak level (about 35% utilization and above), network connections are dropped. This is the first manifestation and usually results in an Abort, Retry, Fail? message for DOS users, and  a "Network Drive Unavailable" message in File Manager for Windows users. The network drives literally blink on and off.

In the Lan Manager days, a second manifestation of peak activity was at login, a message appeared to the effect of: "You have been logged onto the domain, but have not been validated by a server". A network drive connection is dropped, login scripts don't execute, then the net drive may mysteriously reappear.

A third manifestation I've seen in most Microsoft networking products. At peak periods, when a large XCOPY is performed to a network drive, or a large copy is performed from File Manager, all appears to complete successfully, only to find out the very alarming fact that there are many missing files and missing directories on the destination network drive. No errors are reported during the copy process.

Another manifestation of this same problem has been seen during program compilation to a network drive. In the middle of the compilation, the network drive connection is dropped, resulting in an Abort, Retry, Fail?, killing the compile, but to have the drive suddenly reappear.

I've seen these and other related manifestations of dropped network drive connections on many networks, some with ultra-clean brand-new topologies. This problem has been following me around for about 6 years and seems to have only a few commonalities between the many experiences:

Microsoft Networking (SMB)
Dropped connections at about 35% and above network utilization.

What I am nearly certain is not the cause of the problem:

1. network protocols and configuration thereof
2. anything topologically related
3. network cards or drivers

All these years of this problem resurfacing has led me to the semi-conclusion that Microsoft networkring is inherently weak and unreliable. Help prove me wrong.

If anyone of the experts out there has seen this problem and has a cure, I owe you dinner--big time.

I'd like to find out the root cause of this continuing problem. Have at it.

Answer : Dropped Network Connections

You have not stated whether your are running a contention based network (ethernet).  As I'm sure you know, sustained network utilizations for 40% or greater render an ethenret entwork unuseable to to the snowballing of collisions and rebroadcasts.  You are dangerously close to this utilization level.  I have experienced similar probles with Microsoft networks during sustained periods of high utilization.  My solution has been to resegment my network so that I keep my average utilization  low.  Either breaking your network up into small segments each passing thru a 100MB Switch or by adding additional NICs in your servers and breaking your network up into different segments will help greatly.  

I know this is not the answer you hoped to fined, but it is a viable solution - as is switching to a deterministic network topology like 100vg.