network back bone
i know this might sound stupid but i dont know were to start..but my manager is thinking about upgrading the network backbone cus at the moment we are only runing 1gbps across the school and our main switch which 4108gl. were would i start i mean how do i pumb up the back bone ( i know it sounds stupid but never actual know how too first time for this) would it be case of finding out what the main switch is cable of and also check the modules inside it and then check all the switchs across the board most of them are the same but if they cant cope the check to see if they can have a module or somthing ..
The following would need to be looked at to see if they could support what you want:
a) The fibre cables - are they OM3? If not, they probably won't support 10Gb, so would need replacing.
b) The edge switches - are there 10Gb modules available for them? If not, you will need new switches at the edge - something like the HP Procurve 2900 series or the new 3500yl series would do the trick
c) The core switch doesn't seem to have any 10Gb modules available from HP, so you'd need to upgrade. Something like the 5412zl would do.
Even with a small network, you are looking at £10k for the core, and £3k per edge switch, plus any cabling. Our school would cost us in excess of £50k to replace our fibres, switches and core with 10Gb compatible ones.
Oh, and I forgot to say - we only recently upgraded from 100Mbit to 1Gbit backbone links (well, we had 1 Gbit link and 10 100mbit links). 1Gbit is standard stuff at the moment. Before we upgrade to 10Gbit, I will be bonding spare fibres around the school to make 2Gbit links to our edge switches (I can't justify that cost at the moment either, as it would mean buying 22 mini-gbics, a new module for the core and 22 patch cables - so about £6k).
1Gbps is (correct me if I'm wrong) pretty common in schools, it's what we use as a backbone across the site.
1) Find out whether your cable will handle more than 1Gbps. If it's cat5 or cat5e, you're on tricky ground. If it's fibre you should be ok.
2) Find out what your switches will handle and upgrade appropriately. Both ends must support the speed that you want to achieve.
3) Persuade your finance department that you're doing something worthwhile even though they can't see it.
If you've got multiple cables you could also look at bonding them, so both cables are used as one link between two switches.
Edit: If I thought less and typed faster, then that wouldn't happen.
The 4108gl has a 96gb backbone across the actuals switch. You do not say if you are one site, or two, or more. If you are linking between sites then you may something like:
4108gl <------ fibre Link ---> 4108gl
and it may be an idea to have a 10gb fibre here?
On the other hand you may have:
4108gl ---> lots of 1gb fibre/copper connections -> lots of satellite switches
Which is what we have here. Unless you have very large amounts of data (video, large pictures, 100s of clients) coming from one of the satellite switches, you probably don't need to upgrade these connections.
And there is the question of if the satellite switched would support faster connections. You'd almost certainly need new fibre modules in both the 4108gl and the satellite switches, and new cable runs to support the faster speeds.
If the satellites are all managed switched, a better idea might be to turn on the 'Spanning Tree Protocal' on all the switches and then invest in cross connection between the satellite switches, creating a Mesh network rather than a Tree network topology.
This won't give you any speed increase as the new connections would be totally unused, but would give you increased resilience/reliability if one of the switches failed.
Ultimately though, it all depends on your current topology, data usage over the connections, what types of switches you are using and where, etc.
The 4108gl doesn't support 10gbit modules according to the HP site. And only has a 36.6GB backplane.
Originally Posted by tmcd35
A cheap option might be to use the port-aggregation features of your switches and have four copper links between each of them. Check before you do this - if have some kind of network warranty then you'll probably need to get some certified chappy in to run bits of cable for you.
Originally Posted by kevin_lane
Unless you're running Gigabit to the desktop, 10GbE probably isn't worth the expense. I would also check just what percentage of the 1Gb backbone is being used before upgrading (if you haven't already done so).
mind over load
lol mind over load
well we have 4 sites 1 of them being the main T block and most of our switches are Hp procurve 2650s and we have a few 2524 and 1 2312 all of this fiber bonding stuff is new ( how does that work) not sure about ethernet cable if its e6 but i know we have brought cable in the past and they have been cat6e. our school is a ict spectialist and we have alot media stuff etc..sounds like alot of work and alot of problems look like the could occur what about down time ?? is it best to do it bit by bit or in the 6 weeks hols ?
Trust me, do it in the holidays.
Bonding just means that both switches know they have two cables between them and can therefore send twice the data to each other.
Graph your actual throughput at the moment for a few weeks, you'll probably find you're nowhere near 1Gbps consistently.
To be fair, I didn't check anything. Just went from memory. I could have swore it had a 96Gb backplane. Still very doubtful even a full 36Gb is being used here..
Originally Posted by localzuk
As for the 10Gb. Like I said, didn't check but it does support channel bonding as other have pointed out.
It's best to do what Michael says and work out if you need to do it!
Originally Posted by kevin_lane
What bandwidth are you currently using? If you don't know that, then measure it (mrtg will do it for free but your switch may well have come with tools which can also do it). When you know what your peak usage is then you can start working out what (if anything!) you need to upgrade.
What connection to you have into your main server(s)? If that's only 1Gbit then it doesn't really matter what you do elsewhere, it's likely that that will be the bottleneck.
What speed are the discs? What are the queues like for discs? (perfmon will tell you) - if you've got a queue of any length then adding network bandwidth is likely to just increase the queue rather than improving user experience!