How good/fast is your network?

[SimpleSi]

I've a few simple HP Proliant machines and a few normal desktops running as file/print/dns/printer servers in my primary schools which seem to do the job of normal worst case of 35 netbooks wanting to save a word file at the same time

 

...but maybe I need to wire the server up using a 100GB Fibre Card with a SCSI SAN FlashRAM super duper thingy and see them fly!

 

But ,my question is - do you big boys and girls measure the traffic on your servers or the speed of data/app load save times or do you just buy as fast as you can afford and assume it miles better than a 10BaseT connection to a Pentium II running NEtware?

 

If you do measure - what with what dear henry - with what?

 

Simon

[CyberNerd]

 

If you do measure - what with what dear henry - with what?

 

 

measure the baseline! i.e. how the network looks with NO (edit: normal would be a better description) load. Then compare that to a loaded network to see signs of trouble.

Other than that I consider 70% load as 'full' and buy the best I can afford.

Edited March 15, 2012 by CyberNerd

[SimpleSi]

So - with what dear henry, do you measure your normal and 70% loads - temp gauge on the CPU?, stopwatch to see how long office takes to load?

 

Si

[CyberNerd]

Solarwinds is pretty good. Nagios does a good job too but It rather depends on the switches that you have and their capabilities.

One thing you will be able to do is mirror a port and have wireshark capture packets, which you can glean plenty of information from. You'd need to do this at strategic parts of the network (the core and near edge switches)

[CyberNerd]

last time I baselined I essentially did this: http://wiki.wireshark.org/KnownBugs/OutOfMemory?action=AttachFile&do=get&target=Using+Wireshark+to+Create+Network-Usage+Baselines.pdf

 

I'm sure there are better tools out there, but it will depend on what your switches are capable of.

[featured_spectre]

Home network is fibre to all devices except laptop and Xbox....though I will get media converters for that soon

[mthomas08]

We dont really measure anything, with netsupport we can tell straight away whos on cable/wireless by the response time of it. Also programs like SIMs can go slower on wireless.

 

We are 90% GB network here and some spots of 100MB's. Seem to be doing alright, we should be fully GB within the next 2 Years but again we seem to be handling alright and we do rely heavily on IT/Network servers and software.

 

Dont really measure it but sometimes may check switches to see how they are coping.

[tbutcher]

Home network is fibre to all devices except laptop and Xbox....though I will get media converters for that soon

 

that seems a little unnecessary....

[winng]

We use wireshark, nagios, observium, spiceworks, etc. to see what is happening across the network. In terms of checking speed of the network, I don't think you can beat using a stopwatch to time a user experience, i.e. logging on to the network and opening a word document! We do this across different parts of the school site and using wireless and wired connections on the same device.

[SimpleSi]

that seems a little unnecessary....

 

But you've got to give him EduGeek credit

 

Si

[SimpleSi]

My point is that lots of people give advice on using X tech against Y tech (ISCSI vs SAN vs Fibre Channel etc etc 100000Gps to each desktop ) but I'm doubting very much if they've ever actually compared a like for like setup and know what the differences are (and whether they are truely significant )

[mthomas08]

My point is that lots of people give advice on using X tech against Y tech (ISCSI vs SAN vs Fibre Channel etc etc 100000Gps to each desktop ) but I'm doubting very much if they've ever actually compared a like for like setup and know what the differences are (and whether they are truely significant )

 

Agreed

 

For us we wanted to upgrade to 1GB and Fibres to blocks. This is to give us breathing room and as time went by programs did start to slow down on 100MB connetions (Aka SIMs) on all the machines that got upgraded to 1GB SIMs ran much faster.

 

Fibre for blocks to me is a no brainer just to keep things going well, we prefer to upgrade ahead of problems but in all honesty I dont think we will consider doing more then a 1GB network with Fibres to blocks for many years to come.

 

With all this so called cloud business which is apparently already here (Actually some time to come yet) we may have to consider paying for more internet bandwidth but other than that...

[3s-gtech]

I migrated our last few 100MB fibre links to 1GB a few years ago, more out of a desire to upgrade than an analysis of load on those links, and I had the parts available to do it for free. I do try to keep a watch on general performance over the entire network, find how things are running etc - word of mouth from the users and the general responsiveness and 'feel' of a workstation's performance on the network can help. I also look at the utilisation of my core switches to see if there are problems there - my core has a monitoring section in the GUI.

 

Rather than moving to 10GB in the near future, I will probably look at providing more dedicated 1GB links. Some are fed into switches via the fibre links (HP Procurve stuff) and the second link feeds another block further down the line, which I'm not massively fond of. Saved running separate fibres at the time I guess!

[SimpleSi]

I'm just wondering if anyone has seen something that shows what sort beneifts/bottlenecks each part of a network provides.

 

e.g (Very simple one )

 

30 Win7 laptops running into an 11g AP into 100MBs Switch connected to a XEON Quad 8GB fast server.

 

Replace switch or AP?

What sort of performance benefit could be acheived?

 

Are their tools for predicting this sort of thing or is it still up to us geek gods to decide based on our knowledge/guess?

 

Si

 

I've seen stuff from Ruckus that suggests that due to all the wireless stuff going on - the 100MBs

[SimpleSi]

word of mouth from the users and

 

Now you having a laugh

 

Who'd trust a user report on how fast things are

 

Si

[localzuk]

We measure network utilisation 24/7 using Cacti. I also use performance monitor on the file server to see how things are going at random times, and at the moment I'm running the Microsoft Assessment and Planning Toolkit to build up a picture of server utilisation.

[3s-gtech]

Now you having a laugh

 

Who'd trust a user report on how fast things are

 

Si

 

To be fair, they're using PCs in other parts of the school where I may rarely go - I've picked up failing switches due to user reports of slowness before now

 

Some of the monitoring solutions suggested look good though.

[pcstru]

30 Win7 laptops running into an 11g AP into 100MBs Switch connected to a XEON Quad 8GB fast server.

 

Replace switch or AP?

What sort of performance benefit could be acheived?

Are their tools for predicting this sort of thing or is it still up to us geek gods to decide based on our knowledge/guess?

I've seen stuff from Ruckus that suggests that due to all the wireless stuff going on - the 100MBs

 

Depends what the clients are doing and how they are set up. In that scenario I'd probably want to add wireless channels so effectively more radios on the access point (or more access points). I'd measure performance from the perspective of the PC's and I'd attempt to validate my hypothesis by first using all the laptops and then ... switching half off them off to simulate the availability of another channel. If this was a site wide thing I might invest in something like AirMagnet if I can't find a free analysis tool that will let me look at contention issues. That starting point is just a hunch (experience) and maybe I would have to look elsewhere if the evidence doesn't support the hypothesis.

 

We use all sorts for monitoring. Nagios, wireshark, Procurve Manager, Widows performance monitoring tools, VMware's tools, anecdotal user experience, a stopwatch. It really depends on what we are looking at.

 

If someone is asking me for money to buy some new piece of wizzotech then I ask for evidence of the current situation and a hypothesis of how their strategy will improve matters. Often enough they have been surprised by what they find and had to rethink what the actual problem might be.

 

Where this is particularly difficult is strategic thinking. We are investing to upgrade our core to edge links from 1G to 10G. It's difficult to bring good, hard evidence to the table to say that this will be necessary or when. Seems a 'no brainier' that the number of devices will increase and that bandwidth demands will increase - but perhaps we will put in the links and everyone will be wanting access via 4G, so our 10G capability from edge to core will never be fully utilised. Ho humm.

[CHR1S]

Home network is fibre to all devices except laptop and Xbox....though I will get media converters for that soon

 

Lol wut?

 

Tell me how your fibre to all your devices is any better than my TP-Link TL-SG1008D 8 Port Gigabit Switch | TL-SG1008D + Cat6 to my devices?

 

After that tell me how much more your implementation cost.

 

You have tweeked my curiosity