Jump to content

Recommended Posts

Posted

The techies have just re-imaged a couple of computer rooms - we had managed to image half the PC's in batches of three in my office which has 1 Gb connections to the core switch.

 

The two rooms in question only have 100 Mb switches and are connected back to the core using fibre.

 

As it's been activites week this week, they decided to re-image the remaining half of each room in situ.

 

One of the rooms imaged fine but they reported the other incredibly slow....

 

What's the best way to test the network speed and then troubleshoot?

 

I was thinking of just finding a large file and timing how long it takes to copy from one of the servers down to a laptop plugged in to each room....

Posted
Even if the switches are 100Mbps, I presume the fibre uplink is gigabit? The best you can expect is around 12MB/s, whereas gigabit will do up to 120MB/s, a huge difference. Most new machines will do in the region of 30MB/s to 75MB/s+ easily exceeding what 10/100 can offer.
Posted
Even if the switches are 100Mbps, I presume the fibre uplink is gigabit? The best you can expect is around 12MB/s, whereas gigabit will do up to 120MB/s, a huge difference. Most new machines will do in the region of 30MB/s to 75MB/s+ easily exceeding what 10/100 can offer.

 

Hi, yes the edge switches are connected back to our HP core switch on 1Gb uplinks.

 

I've just tested a laptop in my office using iPerf and got from the laptop on our wifi to my desktop PC - 70 Mbits/sec

 

Going to test the two different classrooms later when school finishes.

Posted
Hi, yes the edge switches are connected back to our HP core switch on 1Gb uplinks.

 

I've just tested a laptop in my office using iPerf and got from the laptop on our wifi to my desktop PC - 70 Mbits/sec

 

Going to test the two different classrooms later when school finishes.

 

That's probably about right - 7MB/s to 8MB/s - I'm guessing the wireless notebook is connecting at 65-130Mbps to achieve real transfer speeds like that.

 

The bottleneck unfortunately will be the 10/100, rather than your uplink. Doing fewer at a time can speed things up - I guess it depends also how big your image is.

Posted
That's probably about right - 7MB/s to 8MB/s - I'm guessing the wireless notebook is connecting at 65-130Mbps to achieve real transfer speeds like that.

 

The bottleneck unfortunately will be the 10/100, rather than your uplink. Doing fewer at a time can speed things up - I guess it depends also how big your image is.

 

Quick test in a couple of IT rooms that were empty:

 

Room 1 : 91.2 Mbits/sec

 

Room 2 : 91.2 Mbits/sec

 

Room 3 : (the one the techies complained imaged slower) = 88 Mbits/sec

 

So whilst a little slower it still seems OK to me....

 

Room 1 and 3 have the same HP 2530-48 switches.

Room 2 has older Allied Telesis 8000S switches.

Posted
All those numbers look about right. I'm pretty sure on a gigabit connection (when imaging), it'll probably use in the region of 300-600Mbps, being limited by the SATA controller or hard disk rather than the network connection.
Posted

Have you worked out the contention ratio of the switches? Because if you are fully loading a 48 port switch which has a single Fibre link with a maximum connection speed of 1Gbps, then your limiting factor is going to be the share of both the uplink and the backplane of the switches in question. At best the switch would be giving 2.6MBps on full load and obviously if the switches are in anyway stacked or worse chained off each other then you are going to see even less, and this is not taking in to consideration the theoretical backplane of your HP switches.

 

With iPerf its best to test comms from a pc to a pc both ways during low and high load periods and see the difference, I work to deliver 200 MBps to all workstations attached to any stack on full load as any real time applications will really suffer on a LAN that can’t offer this.

 

Also what are you using to do your imaging?

Is in multicast enabled? and are the switches?

 

Out of intrest do you have;

A toplogy diagram for your network

HP procurve manaager or Solarwinds Orion or simular?

Posted
Have you worked out the contention ratio of the switches? Because if you are fully loading a 48 port switch which has a single Fibre link with a maximum connection speed of 1Gbps, then your limiting factor is going to be the share of both the uplink and the backplane of the switches in question. At best the switch would be giving 2.6MBps on full load and obviously if the switches are in anyway stacked or worse chained off each other then you are going to see even less, and this is not taking in to consideration the theoretical backplane of your HP switches.

 

With iPerf its best to test comms from a pc to a pc both ways during low and high load periods and see the difference, I work to deliver 200 MBps to all workstations attached to any stack on full load as any real time applications will really suffer on a LAN that can’t offer this.

 

Also what are you using to do your imaging?

Is in multicast enabled? and are the switches?

 

Out of intrest do you have;

A toplogy diagram for your network

HP procurve manaager or Solarwinds Orion or simular?

 

Thanks for helping.

 

I'm trying to understand why the techies did exactly (roughly) the same operation to a handful of PC's in two different rooms yet only mentioned to me that ONE of the rooms seemed a lot slower to image and install software such as CS6.

 

Imaging is done using WDS.

 

Each edge switch has a fibre link to it - there is no stacking.

 

I've just plugged a laptop in to each switch in the area I'm investigating and can't see any issues on the switches.

 

Did notice a couple did nt' have the default gateway set but can't see that being the issue.....as I'm sure it will be same in the other room - the switches were put in at the same time...so it's been missed from all of them I imagine.

Posted

If you are imaging a whole room with WDS, then you should enable the multicast mode (as long as it isn't disabled switch side)

 

What you are probably finding is that they are all going as Unicast sessions at the moment, so that the server is reading the files lots of times which will slow down imaging dramatically.

 

I would avoid running a multicast during a working day, but it should speed it up a bit and allow you to image them all at the same time

Posted

Last year I replaced all the edge devices onsite at our school to be Gigabit, but this shifts the bottleneck upstream

 

If you are using MDT to deploy your computers you can use Multicast to deploy your image, i am not sure it helps much with the 10/100 bottleneck though.

  • Thanks 1
Posted
Last year I replaced all the edge devices onsite at our school to be Gigabit, but this shifts the bottleneck upstream

 

If you are using MDT to deploy your computers you can use Multicast to deploy your image, i am not sure it helps much with the 10/100 bottleneck though.

 

I think we're better off bringing the PC's back to my office where we can image 12 at a time - all connected to the core switch.....the techies were trying to save time unplugging and moving - but it actually took them longer....!

Posted

I guess there could be a number of factors - file transfers, users logging on, internet downloads - all eat into available bandwidth.

 

Long term upgrading to gigabit to the edge will make imaging noticeably quicker and is well worth the upgrade.

 

It's yet to be seen whether 10Gbps will ever be squeezed out of CAT5 without some major technical development.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



×
×
  • Create New...