Network Delay

  • Thread starter Thread starter Guest
  • Start date Start date
G

Guest

I have a windows network of 4 sites connected by a At&T T1 frame. I have
768Kb connection to each site and can max the pipe out by moving large files
across it. However, when I do simple things like try and see the contents of
a folder on a remote 2003 server or highlight a remote file so I can
right-click and get properties things will lock up for 30-45 seconds and
progress very slowly. This happens at all 4 sites on all the computers.

Any ideas why simple navigation is so incredibly slow?
 
You are a little light on information to really help debug the problem - so
instead I will give you some starting points and we can go from there.
If you can run a quick sniffer on the host you are experiencing delays it
might allow you to determine what is taking so long (client issue - server
issue - network issue). Next folks are going to need to know:
Is this an AD configuration - if so - how is it set up given your network
configuration?
Are you running AD integrated DNS?
Do you have WINS running for legacy reasons?
Since you are doing Frame-Relay - what is your CIR?
What sort of network equipment do you have set up between the devices -
firewalls, proxies, etc?
What OS version are the clients running?

We might be able to help more with some more specific info.

Regards,
Ed Horley
Microsoft MVP Server-Networking
 
I am going to load a sniffer app (Ethereal) on a laptop but am not really
sure how to use it once I have it loaded. I will need to familiarize myself
with this process.

As for the other questions:
Is this an AD configuration - if so - how is it set up given your network >configuration?
Click to expand...
AD with a domain controller/GC at each location, each location is a seperate
Site and each of the 3 satellite offices have an IP Site-Link to the primary
office. The primary office DC/GC holds all FSMO roles. AD database is 58MB.
Are you running AD integrated DNS?
Click to expand...
non-integrated DNS. Primary records held in primary office DC/GC/DNS box.
secondary DNS server in primary office and one in each satellite (same as
DC/GC) are cacheing only.
Do you have WINS running for legacy reasons?
Click to expand...
WINS running (for no good reason) on primary office DC/GC
Since you are doing Frame-Relay - what is your CIR?
Click to expand...
CIR in main office 1.54Mb, satellite 768Kb each burstable to full 1.54
What sort of network equipment do you have set up between the devices -
firewalls, proxies, etc?
Click to expand...
everything is enclosed in perimeter firewalls but nothing between
sites/offices. Cisco/Alcatel Gb+ capable switching equipment with Cisco
routers and CSU/DSU at either end of Frame connections.
What OS version are the clients running?
Click to expand...
Clients all run 2000pro, XPpro, Servers 2003

My question: If I file copy from primary office to a satellite office I can
saturate the link (I see 1.3 -1.35Mb because of the satellite CIR's burst
capability). But if I browse the directory structure in the same file server
it can take as long as 30 seconds to expand a folder contents even if there
are minimal folder contents. And opening a file over the network takes much
longer than 1.3Mb should allow.
We are monitoring bandwidth usage of the frame at the primary site with MRTG
and barely see 6% usage anytime/ever except during file copies. Yet users at
satellites complain extensively of slow network performance in
viewing/opening files.

could this issue be helped by:
A. active directory enable DNS ( We had this config but changed due to a now
resolved AD propagation issue - I don't think things slowed down noticeably
after the change so that might answer this question)
B. Having only one AD site instead of 4 (but wouldn't this increase network
traffic - again maybe a mute question)
 
Since file transfers seem to be working ok but there is a problem browsing
network file shares and you are running AD in non-integrated DNS mode I
would start there.

Check to make sure your SRV records are making it into your DNS server for
each of the file server. Make sure that all your secondary DNS servers are
getting the SRV records properly from the primary. If that is not happening
that could cause some delay (I doubt 30 seconds though).

Are all your hosts doing DHCP? Is the DHCP server doing dynamic updates
into your DNS server correctly, reverse records making it into the mix also?
Oh - I am assuming we are routing on your network and it isn't bridged so
you either have multiple DHCP servers each managing a local scope or a
single at the hub site that has multiple scopes and you are using
dhcp-helper on the routers.

Also, make sure all the local hosts on remote subnets are doing queries
against the right AD server (which I am assuming for the remote locations is
doing DNS also). You might have to modify the local scope properties to
make that happen if you are using a central DHCP server. Make sure you
local client machines have all the correct configuration information for
their given subnets. Do they have your internal DNS servers entered first?
Not having your internal DNS servers first could cause the sort of delay
problems you are indicating. You would have to wait for a timeout while the
public DNS server search for something they cannot find.

You shouldn't need WINS at all give the client and server OS's you have so
you might want to shut it off as a test to see if that is effecting
anything.

I would think an AD integrated DNS would be easier for your environment and
if you can go to that without major headache do so. Also, 1 or 4 AD sites
is a non issue.

Hope that helps.
Regards,
Ed Horley
Microsoft MVP Server-Networking
 
Ed,

Thanks for your insight. It has given me some things to consider. Hopefully
I can knock this dead.

One suggestion I received from a colleague was increasing the TCP frame size
or to switch to UDP transport. So I will see if those suggesztions have merit
as well.

Will
 
Will,
Interesting, I doubt it is the TCP window size or a difference of transport
protocol. I have a feeling it is a lookup problem that is causing the issue
due the amount of time it is taking. Most likely something is having to
timeout prior to the next lookup starting. At least that is what my
instincts are telling me. When you do large file transfers, are you doing
them by IP address, FQDN, or hostname? Are you using FTP to do the
transfers or something else (SCP, tftp, etc)?

Regards,
Ed
 
Back
Top