Intermittant network disconnection

[Guest]

I am a user on a large network. On this network we have several shared
resources that are constantly available to us. In addition, our machine
client (user) configuration(s) are also located on a external server, not the
local machine.

Several times a day something odd occures. The network appears to 'hiccup'.
Meaning that the shares (shared drives and resources) appear to disconnect
for a very brief period, then re-establish their connection. The visual clue
to this is that small red "x's" appear on the shared resources (drives). Also
the desktop icons seem to disappear for a second, then re-appear.

Additionally, if a file on a shared (remote) resource is being worked on,
very occasionally becomes corrupted. (I assume the file is being written to
while a network break occures (briefly) during the write or save process.

I have seen these symptoms in two totally different organizations, each
using larhe LANs with multiple subnets.

Any clue as to what is going on?

 

[codigo]

I am a user on a large network. On this network we have several shared
resources that are constantly available to us. In addition, our machine
client (user) configuration(s) are also located on a external server, not the
local machine.

Several times a day something odd occures. The network appears to 'hiccup'.
Meaning that the shares (shared drives and resources) appear to disconnect
for a very brief period, then re-establish their connection. The visual clue
to this is that small red "x's" appear on the shared resources (drives). Also
the desktop icons seem to disappear for a second, then re-appear.

Additionally, if a file on a shared (remote) resource is being worked on,
very occasionally becomes corrupted. (I assume the file is being written to
while a network break occures (briefly) during the write or save process.

I have seen these symptoms in two totally different organizations, each
using larhe LANs with multiple subnets.

Any clue as to what is going on?

Verify any multi-homed servers that are acting as routers to subnet the
network. Is a default gateway supplied at the server for each NIC? If so,
try setting the same default gateway on both Nics. Examine the activity on
the subnets, get a base performance reading and determine if too many
clients are residing on the same subnet (too many collisions per subnet).
Consider adding DCs to provide localized shares and logon facility. Examine
what logon server the clients are actually using. Is the Boston client
logging into the PDC emulator in LA instead of the local DC? Is the domain
partitioned into Sites so as to control replication and local logins?

 

[Guest]

Thank you for the clues as to what might be happening. I suspect the problem
is there are too many machines on a single subnet. The problem appears to
happen only during peak work hours between 9:30 am and 2:30 pm.

The base performance tends to peak during these periods (and there is a lot
of heavy network use during this period).

I'll have to examine the domain partitioning. On the first network I noticed
this happening there was no domain partitioning, however the organization
only had a few hundred clients. The present organization I am observing this
has well over 10,000, so I will need to have a closer look.

Thank you again.

Tim Kern

 

[Phillip Windell]

Thank you for the clues as to what might be happening. I suspect the problem
is there are too many machines on a single subnet. The problem appears to
happen only during peak work hours between 9:30 am and 2:30 pm.

The base performance tends to peak during these periods (and there is a lot
of heavy network use during this period).

I'll have to examine the domain partitioning. On the first network I noticed
this happening there was no domain partitioning, however the organization
only had a few hundred clients. The present organization I am observing this
has well over 10,000, so I will need to have a closer look.

The number of Hosts per subnet should kept below 250. The 24bit mask that
gives 254 hosts persubnet is the perfect size. Once you get above 250 -
300 hosts the Ethernet effieciency starts to degrade.

 

[codigo]

Thank you for the clues as to what might be happening. I suspect the problem
is there are too many machines on a single subnet. The problem appears to
happen only during peak work hours between 9:30 am and 2:30 pm.

The base performance tends to peak during these periods (and there is a lot
of heavy network use during this period).

I'll have to examine the domain partitioning. On the first network I noticed
this happening there was no domain partitioning, however the organization
only had a few hundred clients. The present organization I am observing this
has well over 10,000, so I will need to have a closer look.

Thank you again.

Tim Kern

Glad to help.

Looks like you are having a physical issue, not neccessarily a logical
issue.
Partitioning the domain, a logical modification, won't alleviate the level
of collisions on the overloaded subnet(s). Unless replication is the guilty
party, i'ld consider a physical modification first.

Perhaps examining the traffic further might help you determine the exact
nature of the problem.

If you need a subnet calculator in order to submit a proposal:

http://www.subnetmask.info/

 

[Phillip Windell]

Looks like you are having a physical issue, not neccessarily a logical
issue.
Partitioning the domain, a logical modification, won't alleviate the level
of collisions on the overloaded subnet(s). Unless replication is the guilty
party, i'ld consider a physical modification first.
Agreed.

Perhaps examining the traffic further might help you determine the exact
nature of the problem.

If you need a subnet calculator in order to submit a proposal:

http://www.subnetmask.info/

Staying with a 24bit mask will prevent ever needing to use a calculator, and
it will create the perfect host-to-segment ratio.

 

[codigo]

Staying with a 24bit mask will prevent ever needing to use a calculator, and
it will create the perfect host-to-segment ratio.

Thats a fair and wise comment.