Do school's need a 10Gb backbone?

[kennysarmy]

I'm looking at our medium and long-term budget plans.

 

Current network is 1Gb backbone.

Current Core switch is an HP E5406zl(J8697A) installed Feb 2012.

 

We have 17 edge locations and a total of 35 edge switches of varying ages but mostly over 7 years old, of the 35 edge switches within the school, 20 of these can only provide fast ethernet (FE) speeds

 

The school is currently quite traditional with on premise servers, storage and Windows clients, however the plan is to move to fully cloud services within the next 3 years.

 

From initial quotes and investigation works, it's going to cost around £45,000 more to upgrade all our fibre links to 10Gb and to provide switches capable of utilizing the additional bandwidth.

 

Is a 10Gb back-bone worth the investment in a secondary school?

 

Cheers

[RobFuller]

I would say yes, I think these standard the DfE have put together are good and I would use this as your justification to SLT/Board - https://www.gov.uk/guidance/meeting-digital-and-technology-standards-in-schools-and-colleges/switch-standards-for-schools-and-colleges

[mavhc]

When monitoring your current backbone, what's the usage like? If it's maxing out, yes, if not, no

[chazzy2501]

I'd think not, unless you think maybe you'll go to a faster internet. As more stuff get chucked online the local storage needs are diminishing. Maybe a 2.5g backbone? but really you could aggregate the switches links and take an existing 1gb backbone to 2 gig. (as you probably have redundant or unpatched fiber anyhow)

[Michael]

It's a really good question and I'm yet to visit a school that has more than a 1Gb link to the world.

 

Historically it's always been recommended to have a faster LAN than WAN, but if your services are primarily WAN based, then it probably makes little to no difference.

 

I would still recommend OM4 in all new installations, even if you only had 100Mb running over it.

[3s-gtech]

We went to 10Gb backbone throughout in 2020, 1GB to desktop. 1Gb WAN in 2022.

 

For most things, we don’t notice the backbone improvement. We would when imaging, if my MDT server was also on 10Gb but it’s not.

 

The WAN connection was another matter, as we were throttling 100Mb constantly even with traffic management.

 

We are still a very much on-site services setup, but I needed to remove the bottlenecks before any movement on that.

[timbo343]

I'm looking at our medium and long-term budget plans.

 

Current network is 1Gb backbone.

Current Core switch is an HP E5406zl(J8697A) installed Feb 2012.

 

We have 17 edge locations and a total of 35 edge switches of varying ages but mostly over 7 years old, of the 35 edge switches within the school, 20 of these can only provide fast ethernet (FE) speeds

 

The school is currently quite traditional with on premise servers, storage and Windows clients, however the plan is to move to fully cloud services within the next 3 years.

 

From initial quotes and investigation works, it's going to cost around £45,000 more to upgrade all our fibre links to 10Gb and to provide switches capable of utilizing the additional bandwidth.

 

Is a 10Gb back-bone worth the investment in a secondary school?

 

Cheers

 

Your network is about the same size as ours, we have about 7 OM1 lines around school with the rest Cat6, a total of 25 runs - we have around 33 cabs. Some lines are Trunked together to give 2Gb lines.

 

We've had a couple of quotes with some local cabling companies as we are planning upgrading a couple of runs with an 8core (approx 90m) and 16core (approx 180m). For us they are around £2k to £3k - the 180m run is one of the longest runs.

 

Then we need the switches so it's going to be an expensive project.

 

We have a 5406zl2 and supports 10Gb SPF+ modules but looking at the traffic through put on each 1Gb line it's not even touching 50%.

 

I mean 10Gb backbone is all well and good but as you say it is really needed? If you were putting new in then yes a 10Gb backbone would be ideal.

 

Also, those schools who have >1Gb internet lines are you really going to see a huge impact at the end users end?

 

I would assess what you have at the moment, look at the throughputs on your core switch to see where the peaks are and start there.

[Guest Guest]

We upgraded the bulk of over back bone to 10Gbps a few years ago (only did it as our cabling was up to spec and was about an extra £20 per sfp)

 

We haven’t noticed any difference apart from when mass re-imaging or pushing out updates

[Michael]

I should add - I do support schools with 10Gb back bones and also servers with 10Gb NICs. There is a difference, but very marginal, simply because I can max out a 1Gb link.

 

It's still the server itself which is the bottleneck, whether you're imaging or just moving a lot of data in general.

[DrCheese]

We upgraded ours last year - Mainly because our previous core switch was reaching EOL (11 years old)

 

We also have a 2gb up/down Internet connection now (Was only £100 extra a year over our previous 1gb!)

 

Do we notice it? No not really. The only real benefit is that live migration of VM's between servers is super quick now but it was hardly slow before!

 

I think it's more about being ready for the future than anything, especially as the DFE requirements now list feeding Wi-Fi points with Multigb ethernet connections as a requirement for all new builds (& if you've had CTC funding)

[psydii]

We've had a 10Gb backbone for 15 years. It was more important when all our data was on premises.

 

@RobFuller and @Michael Both make excellent points.

 

I would lean towards advising that meeting the DfE spec as the prudent path to follow.

 

However if finances are very tight, and 1Gb has proven more than adequate over the last 10 years, then resiliency and redundancy should be more important than speed of the active kit on the backbone.

 

If one were to draw the physical fibre topology from a logical perspective, we have (in effect) a 10Gb star topology from core to edge and then a 1GB ring between the edge switches and spanning tree keeping it all converged. This topology has saved us many times, and at no point have end users noticed when their link failed over to the 1gb/s ring.

 

One thing 10Gbit does help with is around software deployment, we can be imaging entire IT suites and it doesn't impact the performance of other users.

 

 

As time marches on our of our internet line (1Gb/s) is approaching 75% average utilisation (over 5 minute periods). I can see us noticeably saturating it within a few years... at which point if we can't get more than 1gb to a switch we might find users reporting performance degradation.

 

If you plan to sweat your active infrastructure significantly beyond 5 years, you would probably regret not having more that 1Gb/s between the core and your switches.

[Jcx500]

I would go 10gb at the core layer and also access switches having a 10gb uplink to the core , if budget allows

 

I’ve seen some servers with 10gb links but most just have usual gigabit trunks / bonds split over two switches for redundancy

[psydii]

We've had a 10Gb backbone for 15 years. It was more important when all our data was on premises.

 

 

I would lean towards advising that meeting the DfE spec is the prudent path to follow.

 

However if finances are very tight, and 1Gb has proven more than adequate over the last 10 years, then resiliency and redundancy should be more important than speed of the active kit on the backbone.

 

Riffing off this for a moment, I've just had a thought.

 

If you don't have redundant/diverse physical routes in you topology - spend this money on making them with OM4, but lighting them up with 1Gb/s for now. Then next time you refresh your switch infrastructure (in 5+ years) you can pivot over these "redundant" paths to 10Gb and the legacy backbone becomes the 1Gb/s 'backup path'.

[dapaulio]

Simply I would also say yes as this will future proof your network and from experience it’s harder to get funding for the stuff they don’t see like the infra structure etc

[FragglePete]

I'm upgrading my aging HP ProCurve edge switches at the moment from 1Gb Uplinks and 100Mb to the Desktop, to 10Gb Uplinks and 1Gb to the Desktops. Have noticed an improvement on a few things, but definitely noticeable during MDT. However, did want to keep to Aruba switches but prices of these just started to rocket over the last couple of years and lead times have been horrendous (40 weeks+).

 

Bitten the bullet and now using Ubiquiti Gen2 Pro switches as their stuff availability is much better, but only their PoE stuff - non PoE stuff seems non-existent still. And their 48 Port PoE switches are so deep which is causing a few issues on some of my cabinets around the school. Generally though, they're working well.

 

We have an aging 5406zl as the core, which I'm upgrading the modules with 2nd hand kit to get the 8 x 10Gb modules - this still soldiers on and the modules can be used in the newer refreshed 5406 switch for when I update the core in the future. Still want to keep the core with Aruba.

 

Biggest difference we've notice is specifying a minimum of i5 (or Ryzen 5) instead of i3 that we used to do - initial logons and mapping of drives to OneDrive, Sharepoint, etc just so much quicker. Our internet connection is 1Gbp/1Gbps and some of our servers do go back to the core with 10Gb links.

 

Pete

[3s-gtech]

I should mention another factor with ours - we upgraded to modern CCTV and were able to use our old 1Gb infrastructure (mix of OM1, OM3 and some OM4 fibre, old Cisco SG300 PoE switches and our lovely 5406zl2) to run it completely independently of the main network. I think we had to run one new fibre in total. Had a major bearing on the project cost, which we could offset against the new switches.

 

The advantages of having multiple fibre pairs to cabinets!

 

We also found that for short runs, the OM1 does indeed support 10Gb with acceptable SNR.

[mavhc]

Distinct lack of graphs, you should only be spending £45k if you have data to show it's the most efficient use of the money

[ThatBoringBloke]

Slightly off topic, but have you looked into 'Connect the Classroom funding'?

 

We had our network fully replaced at a cost of £150k all funded by the government.

[kennysarmy]

Slightly off topic, but have you looked into 'Connect the Classroom funding'?

 

We had our network fully replaced at a cost of £150k all funded by the government.

 

Not off topic at all I'd say - but currently Gloucestershire is not included in the funding