Obviously we don't do the "when they break" with servers as these are not lifetime warranty and therefore after 4 years are replaced in the same manor the desktops are (although this is 5 years due to stones 5yr warranty).
With regards to the Core This is replaced every 4/5 years and can't really be pushed out to the edge due to it being larger then most of the edges cabs.
There is generally a reason (increased capacity or throughput requirements) to replace before they break (with an expectancy of 7-8 years). Older switches are moved from the core to the edge. The switches that are taken out are kept for 'Band Aid' solutions until a new switch can be purchased, but the ones on the shelf here are 10 years old, unmanaged and only have a 1Gig throughput.
If there wasn't a good reason to replace switches then why would you until something stopped functioning? - All of the switching kit here has a lifetime warranty.
The bursar here expects to replace the switching over time, but typically the 8 year lifespan and the rate at which IT moves means that whilst you can plan to replace this kit you need to adjust your plans every year.
Everything else is replaced on a cycle (Either 3, 4 or 5 years depending on the kit and warranty!)
Last edited by Mr.Ben; 11th February 2012 at 12:18 PM.
When they break, but we do always have at least 1 switch boxed and brand new on standby, if it gets used another is ordered. Worked well so far.
Good switch gear should last minimum 10 years at which time it should be changed, having said this I have a spare core switch already configured to drop in within 10 mins should ours fail, also have a spare 48 port edge switch to swap out should any edge switch fails.
Other than using lots of cash to create a full mesh topology where if one switch fails others are there to share the load then there is little one can do.
I did have a core switch failure once but had the network up and running within 10 mins due to having a spare which is not as costly as the above method.
As long as switches are kept free from dust and cool they should last a lifetime of the network.
I am thinking of upgrading our infrastructure soon to Juniper but at this moment feel a little reluctant as our Cisco kit is still working very nicely with not only workstations and peripherals running across it but also our CCTV of 30 IP cameras and is only using a third of its bandwith capacity at full use.
If you have plenty of money then I would invest in good quality equipment with a spare core and edge switch change equipment every 10 years accordingly as with the capacity and requirements of the network.
If your equipment supplier has a lifetime warranty then this is based on what the manufacturer of that piece of equipment expects it to last, the price of this equipment reflects this and this is why they can afford to cover it in this way.
Last edited by bossman; 11th February 2012 at 01:38 PM.
It's probably not any sort of accurate timescale from our network - I have updated a huge number of switches from 100Mb/s fibre uplink to 1Gb/s, as well as fitting new devices as they needed to be silent in offices. Very little of the equipment is over 5 years old, so it's difficult to guage lifespans as that's how long I've been here. I suspect I'll move to Gigabit to desktop before much of this stuff breaks!
Yeah we still have 8-10 year old cisco's still in use we have no money. We have a procurve core! I been told I have some budget so I am planning to ditch another load of Cisco for Procurve brilliant kit for the money. Got some 5406's and 5412's to roll out as well.
Apart from a lightning strike (8-10 switches) and a particular switch design by Netgear (48 port Gigabit, should have been actively cooled, wasn't), every network hardware replacement here has been a case of "we need more ports/bandwidth/functionality".
Like you say, things would be better if people did actually put properly specced switches in the first place. But from other posts on edugeek people seem to be quite happy to put in 'enterprise' level software and servers but when it comes to the datacentre things get somewhat neglected. A decent SNMP overview of bandwidth and processor usage on the switch helps to plan upgrades, but I don't think many do this either. it seems to be a put it in and forget about it mentality.
Preaching to the choir. There are threads on this forum about network performance analysis, testing for and diagnosing bottlenecks as well as how to sell hardware costs to SLT in a "benefits to the school/teaching and learning" way. Where the topic coverage is too shallow, there's places like the Server Room at Ars.Like you say, things would be better if people did actually put properly specced switches in the first place. But from other posts on edugeek people seem to be quite happy to put in 'enterprise' level software and servers but when it comes to the datacentre things get somewhat neglected.
For every tech who doesn't bother to read them, that's one more school who might pay my consulting rate and receive the same thing I'd tell them on edugeek for free.
I can see where Localzuk is coming from. I have the same situation, 30+ switches which are approaching 5 years which were installed by a managed service provider. We currently have two physical networks, one for the main school network and one which is solely for FM systems. (Door controllers, BMS etc). Both systems are Cisco. We are planning on replacing our door controllers to a wireless system therefore the FM network is almost redundant so I will move the hard wired stuff over to the main network and VLAN it off. The FM Switches which are slightly lower spec than the main network may work as spares or these can be sold to another organisation and the money used to buy one or two spares for my site. The Data Networks team at the LEA suggested I bought HP Pro Curve switches which will work with the Cisco kit but were half the price so I am looking at both options. Needless to say I have no spares the moment.
There are currently 1 users browsing this thread. (0 members and 1 guests)