1

Yes the age old “the network is running slow” complaint.

Over the last few months I’ve been hearing this but I really can’t pin it down to anything. For example 10 mins ago my boss said that the network is running slow, what he meant was he had Word open and when he saved to his desktop (which syncs/redirects to our file server) it took a while. Others have said they are having slow log on issues first thing in the morning, but not all users. Nothing really stands out for the slow log on users, other than a large amount of documents in their documents folder (but syncing is only carried out at log OFF not log ON).

The only thing that stands out to me is the Spiceworks Network monitor, it consistently reports our switches as having high network bandwidth and it can look huge at times, for example Spiceworks informs me when average bandwidth on a switch is above 100MB and I’m getting reports of anything from 500mbs to 1.5GBs average bandwidth. I even have a report of it hitting 7GB.

The number of staff here has grown over the last year and despite me pushing it, I’ve not really had a say in just where we’re putting staff, in a few locations I’ve had to resort to unmanaged switches being plugged into ports because there are not enough ports for the number of people they want plugging in.

As well as this, our network carries public network wifi traffic.

These switches are HP 100 BASE, only the two core switches are gigabit, which I also don’t think is helping.

Do you think it’s as simple as it looks? The network is just simply overloaded and new switches are needed to cope with the load?

26 Spice ups
2

It sure sounds like your 100Mb/s network is what is killing you. @100Mb/s your theoretical throughput is about 12MB/s, where GbE is 125MB/s. I would imagine that your 100Mb/s uplinks are fully saturated. I would at least inspect the port counters on the managed switches to see if any of your ports are taking errors. I would concentrate on the uplink ports first then the access ports.

One thing you left out is how many users?

12 Spice ups
3

One thing that stands out to me is the public wifi. That should be on a completely different subnet with traffic to the LAN blocked at the router. I’ve also had to do the “switch here, switch there” thing too. I really try to avoid it anymore. With something like this I’d try to start segmenting some things off and see what difference it makes. I also do not use folder redirection. As George said, probably going to need a lot more information to be of any real value. General advice would be to pick an area/department and start there. One last thing, as far as 1Gb/sec Ethernet, you still need the infrastructure to support it.

4 Spice ups
4

100mbps switches don’t really have any place in a modern environment IMO, so I would look at what you could do there especially if the uplink to the “core” switche is only 100mbps.

What are your core switches?

Basically a bit more info on the (rough) topology would be useful i.e. do people connected directly into the GbE switch have problems too?

Also what do you mean by “Public Wi-FI” - dedicated VLAN or AP’s that tunnel back to a central controller?

I’d start with logs if the switches have any, and then with SNMP monitoring of throughput on key ports.

2 Spice ups
5

I’d start by scanning your network with Wireshark and look to make sure nothing is spamming broadcasts or anything.

Other than that it could just be your switches aren’t up to the job.

3 Spice ups
6

Hey guys.

The public wifi goes though our APs which are on our infrastructure. I’ve segregated it in its own VLAN and it also leaves the building down its own ADSL connection separate from the business fiber line. Ultimately though, the public traffic is hitting our switches.

We have around 100 users in the building, give or take depending on the day.

There are some other factors I think as well, general stuff like some of the machines are older than others as well as users talking, ie one user says “I’m running slow” so of course 4 other users sat near them suddenly think they are running slow as well. I’m not saying that they are not, but I think sometimes they are putting ideas into each others heads.

Overall though, I think I need to start pushing for an upgrade of our HP switches, which are about 5 years old now and I’m sure would last longer but we’ve grown and I’ve not been able to grow the network with them.

2 Spice ups
7

You definitely need to look at upgrading those switches to GbE switches, it will make a huge difference. I don’t even use 100Mb/s at home. It may be worth getting switches with more ports if you are having to use desktop switches (if that is the issue and not lack of patch ports).

Explain to management that the old switches are an issue, give them the numbers George1421 gave, its an obvious bottleneck to any computer that has redirected files.

2 Spice ups
8

There are some other factors I think as well, general stuff like some of the machines are older than others as well as users talking, ie one user says “I’m running slow” so of course 4 other users sat near them suddenly think they are running slow as well. I’m not saying that they are not, but I think sometimes they are putting ideas into each others heads.

No! Say it isn’t so. Ha-ha.

4 Spice ups
9

Are all the users that are slow connected to the Fast Ethernet switches? I would definitely go for upgrading to gigabit Ethernet switches for the access side especially if you are syncing profiles all over the place. Are the complaints around the same time of day (i.e morning and evening when users are logging in and out and syncing profiles).

As someone mentioned above, Fast Ethernet switches are dead IMO. But also make sure the computers have gigabit capable network adapters for best performance. if the adapters on the desktops are fast Ethernet, the switch link will negotiate its speed down to Fast Ethernet.

Try and identify any and all patterns. reported times, user location etc. I find that getting this initial info usually helps as a good starting point.

10

Replacing all your switches seems like an easy solution. But, like most easy solutions, it may not help you at all. Before you can fix the problem, you need to identify the problem.

If your switches are 100Mbps, then I’d guess your wiring is as well. So, you’re not going to get >100 to the desktop. If you let the switch negotiate 1G on 100M wires, you’ll be causing other problems.

If your problem is that the link between switches is saturating, then you can look at improving those links to carry more traffic. Can you trunk two lines between switches as a test case? Does the aggregate bandwidth go up?

Is the real problem a bottleneck at the server? Would teaming two NICs improve your throughput there? Is the server in the wrong location? For example, does it sit at the end of a daisy-chain of switches so that every packet passes through every switch (like roads leading to a city in rush hour) or is it centrally located to the traffic pattern?

Unmanaged switches (port-splitters) are fine as a way to expand your network if used intelligently. They give you time to design a better solution.

If you measure your public wifi and find it using too much bandwidth, try throttling it and see if network performance improves.

If you took your car to your mechanic and said, “Sometimes when I press the gas, it hesitates” and his immediate response was, “Well, you need a whole new engine” would that satisfy you?

25 Spice ups
11

+1 to that, no sense in spending money on the hope it will fix it. On a side note what is the wiring in your building? Cat5 Cat5E?

3 Spice ups
12

Verify your cabling and punch-downs will run at gig. Pull some statistics from your uplink ports on your current switches and see if you are coming up against your utilization. That will give you an idea if upgrading your backbone is going to help. Take a good look at your topology and tell your bosses to get ready to spend some $$$.

2 Spice ups
13

Actually, I’d not considered the cable, you’re right we could be only running C5 and not C5E.

I’m going to check a few things. Thanks for the input guys, all of it very useful.

14

A lot of folks are telling you to go to gig switches, and I agree you

will need to do so sooner or later. But you need to survey the land as

well. Does the current cabling support it? Do the NICs in your client

machines support it? Dont assume that just because you bought it in the

last 5 years that it’s a GB NIC.

15

Unless of course that money will help you diagnose the problem. Downfall about some old switches is they have very limited reporting and have a harder time even pinpointing your issue. Modern switches will not only fix one bottleneck but will also give more helpful tools.

I once has the president of the company tell me that he wanted to know exactly what the problem was before he spent money to fix the problem, But he wouldn’t spend the money on the tools to diagnose the problem either because we couldn’t confirm that the tools would help diagnose the actual problem. It was an ugly circle.

That being said, find the switch you want and ask your vendor if you can have a 30 to 60 day trial. Swap it out with of on your 100 MB switches. If it fixed the problem, you know it was your switch. If it doesn’t, use the tools on the switch to help you diagnose the problem. Win-Win.

3 Spice ups
16

Get rid of those 100mb switches they are killing you.

17

The end users that are experiencing the issue how do they connect? Are they using a Cisco IpPhone or Polycom? Are they connecting their workstations through the link on the phone. If so verify what the phone handoff is 100mb or 1gb. Also, those public wifi users are probably part of the problem. Do you have a wireless lan controller? You should manage the public wifi on a separate ssid and vlan, and throttle the bandwidth connection. I would be interested in seeing a drawing of your environment.

18

Spending money to diagnose the issue vs spending money blindly hoping to solve it are 2 very different things :slight_smile:

2 Spice ups
19

As I suggested above, the port counters on the managed switches will give you valuable insight. If your wiring is at fault or there is just something funky with the link, you will see a lot of errors on the ports with questionable wiring.

20

Yes, but even consider 1/3 of the 100 systems on a 12MB/s link (somewhere). It would be nothing for 3 modern computers to saturate that 100 Mb/s link pretty easy. So far I still have my initial reaction to the 100Mb up links causing the slowness.

If I had to spend money wisely, I would upgrade the core switch to GbE then start working down the network tree (away from the core, towards the end users) upgrading the switches.

Just to show the bosses that the money was spent wisely I would benchmark each upgrade pre and post upgrade to show the % of performance improvement / (decrease).

1 Spice up