Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

We have a server running Windows storage server 2003 with: - an FTP server and - VMware Server running a VDS with apache (vds runs windows web 2003)

This is in a data center and we just switched to a new network provider.

Before the switch, things were just fine.

We then switched to a new network provider in the exact same building (there are multiple companies that provide network services in that data center) and things have been very troublesome...

The speeds kept fluctuating and were at first extremely low, like fluctuating between 20 kBps and 100 kBps. Usually more like 20 kBps.

THEN- They said they changed the port speed on their router to auto-negotiate, and that improved things a lot, but we still have serious connection issues.

The speeds are now about 100-200 kBps and some people have no problems at all, but some people have a lot of trouble with speeds, and with the VDS, even some dropped connections.

Another thing is, maybe its just me, but the CPU seems busier than usual.

Any help is appreciated! Thanks!!!

share|improve this question

3 Answers 3

As with any network issue start with layer 1. As suggested above, check duplex settings. Check all your cables for problems using a tester. Verify that you are using the correct pinouts on all your connections. If this is for EoC, make sure all pair lights are lit. If this is a T1 (bonded or single), check error logs for slip seconds, unavailable seconds, and check wan event logs for alarms. Ensure that all interfaces are active if it is bonded. Ping to your DNS and check for jumps in return time (jitter) or packet loss. On a business class level connection (t1, eoc, etc) you should see no more than 2% loss over an hours time when pinging the primary DNS. Have your ISP check your backhaul/trunk for over-utilization, force them to move you if it is more tha 80% utilized. Make sure your own connection is not suffering from over-uutilization, employ some sort of network monitoring using software like spiceworks or wireshark to track if necessary.

share|improve this answer

First thing I would try is setting both to Full Duplex to avoid any negotiation issues. While network hardware should handle negotiation correctly, it can still have issues with some hardware. What do your interface stats look like? collisions? network errors?

CPU load might be higher depending on retransmits, etc.

3 minute hold due to a robotic looking answer.

share|improve this answer
    
Agree with this - set speed and duplex manually - I've seen autonegotiate cause problems when one side thinks the link is Half duplex and the other thinks it is full-duplex. –  Mitch Miller Sep 16 '10 at 5:47
    
How can we determine these interface stats on Windows? We are coders, not network engineers so this is a bit new to us. –  Samuel Sep 16 '10 at 16:00
    
O and also.. We use an unmanaged switch... The setup is... From their Cisco, to our unmanaged –  Samuel Sep 16 '10 at 16:09
    
(hit enter accidentally)..from provider's cisco to our unmanaged switch which is a DLink DGS 1024D, to finally our 3 servers (attached to our dlink).. The dlink has QOS which I fear could be an issue. Also.. Is it at all possible that our bandwidth providers just stink? (and that vmware is a separate issue?) I want to ensure that's not a possibility. The backend bandwidth is Cogent, Bell, Teleglobe and TorIX. And, how can I force port speeds if my switch is unmanaged? Or determine the status? If not possible, do I need to go and buy a new one? –  Samuel Sep 16 '10 at 16:18

Samuel

First - Get a good smart switch. HP gives a lifetime warranty - as does NetGear.
http://h10010.www1.hp.com/wwpc/us/en/sm/WF05a/12883-12883-4172267-4172304-4172281-3963985.html would be a good start and comes in an 8 port and 24 port option.

Second - NEVER AUTO-NEGOTIATE PORTS - always get them to hardcode it and you to do the same.

I trust your in a carrier hotel - so chances are they are simply handing you off Cat5e or similar.

Just to be safe ask to schedule an outage and then schedule them to do a test over that connection.

I have been in some of the Top Carrier Hotels in the nation - and found the strangest things

For example in the Carrier Hotel facility @ 401 N.Broad in Philadelphia we were handed a cat5e cable - two of them. For the life of me I could not get a signal across the wire - no matter what I tried.

The lights would only light up when we had both plugged into the switch - and then seconds later we would have an STP (spanning Tree Protocol) warning kick off.

The Datacenter technician giving us the network drop for our client created a network loop vs doing the wiring correctly.

I then had to scream and pitch another fit when the same character for another client did a bad crimp and he insisted that all of the network gear in the cabinet was bad on our end.

A few letters later he was gone.

That being said - ALWAYS have them check the wire - do a simple iperf if possible to verify the connection can carry what you should have.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.