Electrical Testing – 4th & 6th Jan 2016   Permanent Link to Electrical Testing – 4th & 6th Jan 2016

December 14th, 2015 by Paul Seward

We’ve been informed that some electrical testing will be carried out in January which will disrupt wired and wireless networking in the following locations:

Mon 4th Jan

  • The Hawthornes

Weds 6th Jan

  • 7 Priory Rd
  • Woodland Court
  • 5 Osborne Villas
  • 12 Osborne Villas
  • 16 Osborne Villas
  • 28 St Michaels Park
  • 32 St Michaels Park
  • The Hawthornes

We apologise for the lack of notice, and for the inevitable disruption that this will cause. Unfortunately we don’t yet have any details of how long the disruption will last.

Posted in Problems / Maintenance | Comments Off on Electrical Testing – 4th & 6th Jan 2016

Results – How much of the University has WiFi?   Permanent Link to Results – How much of the University has WiFi?

December 2nd, 2015 by Mark

Following on from my “How much of the university has WiFi” musings, I can now give you a definitive answer – or can I?! Carry on reading if you want the slightly technical details or simply skip ahead to The Results if you want the quick answer!

Our original plan was to combine 3 measurements from which we could generate a quality index. That is to some extent still the goal but it is now in our longer term strategy (or when opportunities arise such as building refurbishments). To achieve a quality network we’d need to invest heavily in design, which is a time-consuming process.

We’ve decided to tackle the problems in two phases:

Phase 1 – Design for Coverage:
In the short term a better option is to simply go for signal coverage in areas that need it urgently. We originally planned to hit a better than -70dBm threshold to indicate good signal coverage but we’ve noticed that the survey hardware has quite good ‘ears’ and can hear our Access Points (APs) much better than your standard smartphone. Because of this we’ve opted for a -65dBm threshold. The plan is to infill areas where we can see poor signal strength and aim to get all buildings up to around 95% coverage. This will be the basis of a bid to senior management.

Phase 2 – Design for Performance:
In the medium to long term we plan to redesign the wireless infrastructure in buildings so that, as well as the all-important signal coverage, we can provide a network that is optimised for capacity and performance that will last for many years to come. Some opportunities to optimise locations will come naturally but we plan to start a programme of improvements in 2017/18.

The Results
We were really quite surprised to find that coverage was a fair bit higher than we’d estimated. The headline figures are:

  • For 2.4GHz we have a coverage of 85% at better than -65dBm
  • For 5GHz* we have a coverage of 78% at better than -65dBm

Our aim for Phase 1 is to expand this coverage to 95% for both 2.4 and 5GHz.

*The expansion of 5GHz coverage is really important because this is where you see a big difference in overall data throughput per user device as a direct result of having more channels available for use.

The Map
The following Google map shows the locations we’ve surveyed and what percentage of that building has acceptable coverage (>-65dBm @ 2.4GHz).

The Google map of coverage from the summer 2015 survey

Green pins indicate locations that have ≥ 90% acceptable signal strength @ 2.4GHz
Yellow pins indicate locations that have ≥ 75% acceptable signal strength @ 2.4GHz
Red pins indicate locations that have < 75% acceptable signal strength @ 2.4GHz

Click on the pins for more detailed information.

The data gathered over the summer vacation has given us all the information we need to improve wireless throughout the university – now all we need to do is get the funding to move forward and we plan to do this over the next three financial years.

Mark Elley
Wireless Service Manager

Posted in Eduroam | Comments Off on Results – How much of the University has WiFi?

Colston Street Fire   Permanent Link to Colston Street Fire

October 12th, 2015 by Mark

Due to the recent fire at the 33 Colston Street residence we are having to change the configuration of the network in the following areas because they are fed from Colston Street:

  • Deans Court
  • Unite Hosue
  • Chantry Court
  • Winkworth House

Work started at 15:55 and we expect no longer than a hour of disruption. Both eduroam and ResNet will be affected during this time.

We apologise for any inconvenience this will be causing.

Update 17:00 12-10-2015
The network changes have been completed and wired and wireless connections are now fully working in Deans, Unite, Chantry and Winkworth.

Posted in Eduroam, Problems / Maintenance | Comments Off on Colston Street Fire

Wireless printers on the UoB wireless network   Permanent Link to Wireless printers on the UoB wireless network

September 28th, 2015 by Mark

It’s not unreasonable to assume that your wireless printer that works flawlessly at home will do the same on the University’s wireless network. Unfortunately it’s not as simple as that.

Home networks vs Enterprise networks
Almost all devices can connect to your home-style WPA2 PSK networks because a passphrase is used to allow the device onto the network. For a device to connect to eduroam it must support WPA2 Enterprise which uses a username and password to access the network. There are no printers that I am aware of on the market that supports WPA2 Enterprise so connection to eduroam is not possible.

Even if you were able to connect your printer there would be an additional problem. On your home network you trust all the people that connect to it because you control who has the passphrase. These people would be trusted to print to your printer. In our environment we have over 40,000 connected devices and you would have no control over who could print to your printer. As fun as it would be to have lolcats popping out of your printer, you’d soon get tired of it!

What about connecting your device to ResNet-Consoles?
I’m afraid this too would be fruitless. The consoles network in on a completely separate network to eduroam, ResNet wired and every other network we provide so you wouldn’t even be able to see your printer let alone print to in wirelessly.

So how can you print to your wireless printer?
In an ideal world we’d have a nice wireless solution to overcome the limitations of your device but I’m afraid we don’t. But, we do have a solution that is a reasonable compromise. If you pop into the IT Service Desk in the Arts and Social Sciences Library on Tyndall Avenue we will supply you, free of charge, a USB cable – it’s a bit old-school but works!

What about UoB printers?
It is possible to print documents on University printers from personal laptops connected to the eduroam wireless service.

Posted in Eduroam | Comments Off on Wireless printers on the UoB wireless network

How much of the university has WiFi?   Permanent Link to How much of the university has WiFi?

August 18th, 2015 by Mark

In the wireless team one of the most awkward questions we get asked is, “How much of the university has WiFi?”. You’d think this would be a question that is easy to answer but the phrase “You know not what you ask” springs to mind.

I can tell you that we have around 400,000sqm of floorspace, 300 buildings, 2400+ Access Points (APs) and that over 40,000 devices use wireless each week… but what does ‘has WiFi’ really mean?

Is signal coverage alone a reasonable metric? What threshold of signal strength is acceptable? Should we be taking noise into account as there is no point having signal when the noise is drowning it out? Are poorly located APs causing problems? Do we have under or over-provision of APs in some areas? Do we have capacity/throughput issues in some densely populated locations? Should we be designing for 2.4GHz or 5GHz coverage?

Unfortunately it’s not possible to answer all these question using data that the APs gather. In fact hardly any of it can.

So how can we work out how much of the University has WiFi? Well, we have to get out there and look, so that’s exactly what we have been doing this summer. We have two teams of casual staff walking around the University taking signal measurements from the 400,000sqm of floorspace using some professional heatmapping software called Ekahau Site Survey. To get around all the buildings it is taking around a man-year of effort, not to mention an awful lot of walking!

So far we have visited over 50% of the buildings but gathering the data is just the beginning. Once the data is gathered we need to analyze it, work which areas are “good” and which are “not so good” so that we can develop a business case to improve and expand wireless coverage where it’s needed.

In my mind there are two elements to the business case; hardware (i.e. more APs) and (re)design manpower (i.e. making best use of new or existing APs through optimum design). The whole point is to get the biggest bang-per-buck by optimising new and existing infrastructure.

We’re gathering data through surveying from which we can generate metrics; a measurement of the quality of our wireless network. Improving the network is an iterative process so any changes we make will require us to resurvey.

Plan A is to use 3 measurements from which we can generate a quality index. The measurements that we feel are important are:

  • Signal Strength – Percentage of floor area that is -70dBm or better.
  • SNR – Percentage of floor area that has 25dB or better Signal to Noise Ratio.
  • CCI – Percentage of floor area that can only ‘hear’ one AP on a single channel (known as Co-Channel Interference).

Combining the above, with a weighting factor, for both 2.4GHz and 5GHz will give us an overall quality rating for a building or area – which we’ll call the “Quality Index”. This will allow us to target hardware or manpower to the places that need it the most.

So let’s have a couple of examples of what results these metrics give us, one for a well designed building and one for a known poorly designed building. I’m not going to mention the names of the buildings as I don’t want to cause alarm! However, the fact is that both buildings work OK at the moment, although Building 2 needs redesigning to optimise it to make sure it can perform well for years to come.

Building 1 – Good signal coverage, good design.

  • Quality index 2.4GHz = 71%
  • Quality index 5GHz = 99%
  • Quality index weighted average = 88%

We will not need to spend any time or effort changing this building as it’s about as good as it’s physically able to get. The 2.4GHz ‘quality’ appears to be quite low so can’t we do something to improve it? Well, the quality is low because of the small number of channels that 2.4GHz has available. There are some changes we could make but the laws of diminishing returns come into play.

Building 2 – Good signal coverage, poor design.

  • Quality index 2.4GHz = 13%
  • Quality index 5GHz = 45%
  • Quality index weighted average = 32%

What’s needed in this building is a complete redesign. The signal coverage is excellent but there are just too many APs. A common misconception with WiFi is ‘more is better’, but this is simply not the case.

When 2 or more APs can hear each other on the same channel they start to chat back and forth to cooperate, and the time they spend doing that is time spent not handling user traffic. So over-provision can end up reducing the effective throughput available to users!

It is likely that one of the APs could be switched off which can actually have the effect of improving performance and certainly reduces replacement costs.

To return to the original dreaded question, “How much of the university has WiFi?”. The simple answer is ‘we don’t know’ but we are getting the data to at least try and answer it!

We plan to build a business case by November 2015 and hope to secure funding for Q1 2016.

Mark Elley
Wireless Service Manager

Posted in Eduroam, General | Comments Off on How much of the university has WiFi?

Issues with beta versions of OSX 10.11, iOS 9 – RESOLVED   Permanent Link to Issues with beta versions of OSX 10.11, iOS 9 – RESOLVED

July 29th, 2015 by Paul Seward

2015-08-03 Update: We think we’ve fixed this now. We’ve had confirmation from one OSX 10.11 user that it’s working and we’re trying to track down an iOS9 user to check.

Technical detail for those who are interested:
Apple appear to be using TLS1.2 (when most vendors using TSL1.0) and the version of freeradius we were using had a few wrinkles in the TLS1.2 implementation. Upgrading to Freeradius 2.2.8 has been on our roadmap since it was released anyway, so we pulled that upgrade forward. That appears to have resolved the problem.

Original Post:
We’ve had a number of users reporting problems connecting to eduroam with the beta versions of OSX 10.11 (El Capitan) for macs, and iOS 9 for iPads/iPhones.

The symptoms as experienced by the user are that the device constantly connects, and disconnects rendering the network connection unusable.

This appears to be a bug in the beta release of these operating systems, rather than a configuration problem at our end.

We’re not the only university seeing the issue, and there’s a lot of discussion happening on mailing lists trying to find a workaround, until Apple acknowledge the problem and release a patched version of the beta.

The wireless team are working with mac specialists inside information services, and at other institutions to try and come up with a solution, but for now our advice is “don’t use beta versions of operating systems to connect to eduroam”

On a similar note, we’ve not had any problems reported to us with Windows 10 – but as it was only released to the public today, we’ve not had a chance to test it yet either! Watch this space.

2015-07-30 Update: We’ve had some information via a mailing list with some suggested fixes. We’re in the middle of a few big bits of work at the moment, but we’ll trial the fixes as soon as we can.

Posted in Eduroam | Comments Off on Issues with beta versions of OSX 10.11, iOS 9 – RESOLVED


IPv6 ready! at 02:34, 10 Jan 21