Small network Big Connection problem

By brush-head? ·
Hi, we're a GP surgery in the UK. Total # of PCs (almost all of which XPSP3) connected 50, in use concurrently averages 30 or so. There is a 2Mbps SDSL link into the NHS N3 network. The majority of PC's are 10/100/1000 Mbps NIC's. We have 4 switches. 1xHP2650 46 ports 10/100Mbps 2x1Gbps, 1xHP1800 24x10/100/1000 Mbps ports, 1xHP2510 24x10/100/1000 Mbps ports. The 4th switch is a Netgear FS524 24x10/100Mbps which plugs from 1 port into the HP2510 & supports 10 PC's almost all of which 10/100/1000 Mbps NIC's. The cables have been added to over the years and as far as I can tell by differing suppliers. No backbone all go to one of the switches above.
Issues I have:-
Overall throughput is not good in my view. A rated 1Gbps NIC even if it connects at 1Gb, has a throughput of 120-140Mbps. Any NIC that drops to 100Mbps has a throughput of 40-60Mbps. Some have problems at boot finding a domain controller, eventually they get a connection most of the time but this can either make the boot time go out to about 6-8 minutes or not at all!
When I've been swapping things around in an attempt to improve things, I get the feeling that cables play a part. Example: PC performs badly in one place. Remove PC take it next to the switch & plug it into a port. Manage to get the higher performance of 1Gb & throughput. Take it back to where it was having made sure the cable to that location is plugged into the port providing the good throughput. When PC located back to original place throughput drops to 100Mb.
The network is lightly loaded. Servers (Windows 2003 Server) are way over specified for what they do so I'm confident the problem doesn't lie there.
Have had PRTG (full version provide by the support Co) running for a few days & nothing suspicious there.
Any ideas out there?TIA.

This conversation is currently closed to new comments.

Thread display: Collapse - | Expand +

All Answers

Collapse -

before you move cables around why not log on

by CG IT In reply to Small network Big Connect ...

to the switches and get their configuration.

That's really the only way to know what might be the problem if there is one.

Collapse -

I already did..

by brush-head? In reply to before you move cables ar ...

Thanks for your reply. That was the first thing I did. There are the odd (maybe 7 in the past couple of months) messages about CRC but no continuing volumes which if there was a problem that was serious I'd have thought that it would generate loads of errors. There are no large amounts of traffic across the network.

Collapse -

How are the switches inter-connected?

by robo_dev In reply to Small network Big Connect ...

How is your LAN connnected to the main network?

Is there a router in between?

My first concern is that if you have >30 users on an ethernet switch with a 1Gig uplink, you are over-subscribing the uplink.

My second guess is that many PCs are mis-negotiating their duplex settings. (connecting at 100mbs half-duplex instead of full duplex). In the web interface of the switch, if you can see CRC-error stats for individual ports, mis-negotiated ports will have millions of CRC errors.

Malware, spyware, and rogue apps (e.g. peer-to-peer apps) can wreak havoc on the LAN....are there good controls in place to prevent this sort of thing?

Is spanning-tree enabled on your network? If there is a topology loop, spanning tree will allow it to exist, but lan performance would not be good.

Without spanning tree enabled, a topology loop will stop the network, period.

Collapse -

Simple LAN...

by brush-head? In reply to How are the switches inte ...

Thanks for your reply.
The cable runs across two floors of the building into the Krone panels of which there are 3. This is a radial arrangement there is no backbone of any kind. The 2 servers are directly connected to auto switching ports which are capable of up to 1Gb each on the main managed switch. The 3 switches are connected with a patch cable to each other e.g. there is a managed switch(this is all auto sensing up to 1Gb), which is connected to a similar but unmanaged switch & from there to another unmanaged switch which is mainly 10/100Mb with 2 1Gb ports which the 2nd switch is connected to. The top floor has a cheap Netgear FS524 10/100Mb 24 port switch which doesn't even have an upload port so one if it's ports are connected to the first managed switch on the floor below.
I haven't noticed any issues with loops but I'll look again. As it is all radial & the traffic is very low, AV scans run nightly and we don't allow introduction of programs into the environment & our e-mail system is fairly tight being centrally run.
Even on machines that lock horns at 1Gb the bandwidth per connection seems to be no more than 130Mb. I wouldn't expect to get the full 1Gb but I don't think it's asking much to get say 200Mb throughput on average.
TIA for any ideas you can put forward.

Collapse -

Unless you explained that incorrectly

by Kenone In reply to Small network Big Connect ...

You have at least one switch that is not gigabit, so no go there. You can only get the top speed of the slowest device that your packets travel through. Wiring counts, if your wiring is old then you may never get the speed your gear can support.

Collapse -

Thats what I think..

by brush-head? In reply to Unless you explained that ...

The wiring is mixed here. Different companies & nasty solutions like using existing wiring & linking to a switch on another floor (not even via an uplink port). Saves cash but poor results.

Collapse -

Hope I understand what you wrote

by mwalsh In reply to Small network Big Connect ...

If i understand correctly it sounds like all your pc's home run into one closet where all these switches are? Is that correct?

If so my first question is with only 50 pc's why so many switches.

Also how are the switches inter connected.

I need to know though if all 50 pc's terminate where these switches are and are the switches in the same closet.

Collapse -

Your'e correct...

by brush-head? In reply to Hope I understand what yo ...

If you look at my reply to robo_dev full details are there. In answer to your question about why so many switches, it's somewhat accidental & not by design.
On our first switch we were about to run out of capacity so a second switch was supplied to cope with a few more connections & to spread the load about. Then the central IT function (who pay for all the kit) decide that they needed to manage all this stuff remotely, put a Linux server in running Novell's Zenworks & an additional switch. Because the practice did not have the top floor of this building until around 4 years ago, I'm not sure if that switch was either there or bought as an easy & cheap way to connect to the existing systems downstairs. That switch could certainly do with replacing as the users connected to it get everything squeezed down a 100Mb link - all 10 of them.
Thanks for your replay & interest.

Related Discussions

Related Forums