WiFi Renewal Project
Rhodes has embarked on a major wireless renewal project that involves the replacement of all existing wireless infrastructure on campus and will see much greater coverage in future. To aid with planning, a comprehensive site survey will be conducted and occupants of affected buildings may need to pay attention to the scheduling of this.
The renewal project will be broken into a number of phases, with the first phase (scheduled for 2013) focusing on major shared academic spaces on campus. The selected solution can ultimately scale to at least 1,000 access points, and in due course the IT Division aims to provide coverage in the common areas of every reasonably-sized building on campus. Individual departments with specific or urgent requirements can fund installations ahead of the central schedule.
- Rationale for renewal
- Enterprise vs consumer wireless
- RFP & selection process
- Site surveys
- Intended vs Incidental Coverage
- Phase 1 rollout
- Phase 2 rollout
- Phase 3 rollout
- Beyond phase 3
- Requests for additional coverage
The University's current enterprise wireless solution was purchased in 2006. Although it's steadily expanded since the initial purchase, it has reached its limits, both in terms of coverage (number of access points) and the number of devices it can support.
The current solution is comparatively expensive, with the installed cost of access points being in the R5,500 range. It takes a one-size-fits-all approach to access points, meaning that the available access points are often significantly better than is required for smaller installations, and that larger installations require many access points. The high captial cost of individual access points (coupled with a restrictive licensing model) has proved a considerable barrier to entry in deploying wireless on campus. There are also significant recurring costs, which mean that it is cheaper to discontinue the current solution.
The wireless technology in use at present is dated, and it no longer supports the latest wireless protocols. Practically, this means reduced speeds, and smaller numbers of client devices. New wireless technologies, on the other hand, provide for much greater client densities, and better speeds. As the use of mobile devices (phones, tablets, etc) grows, better wireless technologies become increasingly important.
The bring-your-own-device (BYOD) trend means that more and more people carry wireless-enabled devices. This is something we'd like to encourage and facilitate, especially in the case of students. Recent surveys suggest that most students carry at least two wireless-enabled devices, yet our Student Network still only provides for one.
In short, we've reached a point where the current solution no longer meets the University's needs and needs replacing.
A question that's commonly asked of us is why we don't make use of the cheap, consumer wireless access points that are widely available. Thus while we're discussing a new enterprise solution, it seems pertinent to address this issue.
Consumer wireless access points (the kind you might use at home or get from your ADSL provider) are designed to support a typical home environment. They typically only support a small number (< 10) of concurrent users, and performance degrades rapidly as the number of users increases. Enterprise APs, on the other hand, are designed for much higher numbers of users (up-to > 300), and do a much better job of ensuring that no single user degrades the experience for others.
Consumer APs have comparitively small coverage areas (a single AP will barely cover the average house), and don't handle roaming (moving from one AP to another) very well. Enterprise wireless usually has much better roaming support (important when you are walking around a larger building), and the indivdually APs typically have better antenna, more powerful radios, and cover larger areas.
Consumer APs typically don't provide any sort of redundancy or resiliancy. When an AP or the underlying network breaks, you simply lose access. Enterprise wireless, on the other hand, is designed to be highly available and tolerant of failures.
From a deployment and management perspective, consumer access points don't scale. Each individual access point needs to be configured manually, which amounts to a lot of man-hours of work when you have several hundred of them. Enterprise wireless, on the other hand, allows us to reconfigure large numbers (up-to a thousand in the proposed solution) simultaniously and within minutes.
Finally, most consumer APs don't support multi-user authentication. Whilst it is fine to have a single shared password for your home wireless network, this doesn't work when the wireless networks supports several thousand users who come and go all the time. Services such as eduroam simply aren't possible on the average consumer access point.
Unfortunately this flexibility comes at a cost. Whilst a home-use AP can be had for a few hundred Rand, an enterprise AP typically costs a couple of thousand Rand. What's considered in this document is an enterprise wireless solution.
During the 2012 budget process, money was allocated from both the IT Division's central budget and from the Student Network to provide for a replacement enterprise wireless solution. These two sources contributed roughly equally, and a total of R2million was set aside for the first phase of a wireless renewal.
In early 2013, a comprehensive 2013 WiFi RFP was sent to the local distributors of selected wireless vendors. The RFP aimed to select a solution that would well exceed our current needs, and ultimately scale to 1000+ access points (i.e. be capable of comprehensive coverage of our campus). Particular attention was paid to the known shortcomings of the current solution.
Whilst not all vendors responded to the RFP, four proposals were received in April 2013. These were evaluated per the criteria in the RFP, with cost and technical suitablity being equally weighted. Two of the proposals were short-listed, and the respective distributors visited our campus to get a first-hand feel for the nature of Rhodes' buildings. Both were also asked to provide sample equipment for a proof-of-concept implementation, and a comprehensive technical evaluation was undertaken. The site visit and feedback from the proof-of-concept resulted in both of the short-listed distributors submitting revised proposals that better suited Rhodes' environment and needs.
The two revised proposals, together with our own experience from the proof-of-concept and information from independent reviews, were then considered to make a final selection. In this regard, the IT Division is grateful to the technical staff in the Departments of Computer Science & Information Systems for the valuable input they provided into this process.
At the conclusion of the process, the proposal by Ruckus Wireless (the vendor) and three6five Systems (their local representative) was selected. An initial order — for two new ZD5000 controllers and approximately 240 access points — was placed on 27 May 2013.
Prior to the roll out of new wireless infrastructure, three6five will be conducting comprehensive wireless site surveys of the largest academic and administrative spaces on campus. In newer buildings, these surveys will be conducted off the building plans. However, in older buildings, a physical survey is necessary. This will entail staff from three6five and the IT Division taking measurements from almost every room within the building in order to work out the optimum placement of wireless access points.
In order of size, the buildings to be surveyed are:
- Main Library (A23, from plans)
- Chemical & Pharmecutical Sciences (C9)
- Biological Sciences (D1, partly from plans)
- Main Admin (A1,A2,A3, partly from plans)
- Eden Grove (B28, from plans)
- Bantu Steve Biko Building (H26, partly from plans)
- Rhodes Theatre (C3)
- Africa Media Matrix (F5, from plans)
- Schonland Buildings (A13,A14)
- Barratt Lecture Complex (H45, from plans)
- Physics & Electronics (A4)
- New Arts (A5)
- Geology (A9)
- Hamilton Building (C10, from plans)
- Geography (A10)
- Struben Building (C7, from plans)
The physical surveys will be conducted over a period of about two weeks, between 24 June and 5 July. As soon as project planning has been completed, the IT Division will make contact with the nominated Departmental Contacts from each of these buildings. In order to ensure a comprehensive survey, arrangements will need to be made to facilitate access to staff offices, common spaces, laboratories, etc.
Where surveys are conducted off-plan, it might still be necessary to validate the design by conducting a physical survey of a small portion of the building. This will almost certainly happen, for instance, in the case of the Main Library.
It is important to note that being selected for a survey does not imply that a particular building will get comprehensive wireless coverage. On the contrary, for the most part these buildings are being surveyed because they're considered complicated and it is impossible for us to predict the cost or infrastructure required for such coverage. Thus the survey results will feed into the planning process for later phases of the renewal project (see below).
Surveys of other buildings
The buildings listed above are certainly not the only ones that will be surveyed; ultimately we'll perform surveys of every building that gets wireless. The difference is that the ones above will be surveyed by three6five, and will thus be on a very tight scheduled. Surveys of smaller buildings will be done in-house, once our staff have been trained.
For interest, the criteria for inclusion in the list of buildings to be surveyed by three6five was academic/admin buildings with a total floor space of more than 1340 square metres. Very simple buildings (for example, The Hanger) were excluded. Struben Building is the smallest that will be surveyed, and two separate surveys of this building will be conducted (one by three6five and one by IT staff). This will allow us to validate our in-house capability to do surveys of other, smaller buildings both as part of this rollout and in future.
After three6five completed their surveys, it became apparent that considerable work would be required in order to get a useable plan for each building. The main reason for this is that the three6five survey did not take into consideration the way particular spaces were used — it assumed that all areas in the building needed the same sort of coverage, and aimed to provide excellent, high-density coverage everywhere. Whilst this might be an ideal situation, it doesn't fit well with the realities of a constrained budget.
As a result, the IT Division staff will re-survey most buildings, using three6five's plans as a starting position. These followup surveys will be more informed about space utilisation, and will make the necessary compromises to try and get maximum coverage out of the budget available without impacting usability in important areas.
In many buildings this renewal project will not provide comprehensive wireless coverage. Centrally- and Student Network- funded coverage will target specific areas, such as public spaces, seminar rooms, etc; this may be supplemented by departmentally funded access points that cover other areas of specific interest to the department concerned.
Access points are deployed so as to ensure that there is good (≥ -75dBm in both 2.4GHz & 5GHz) coverage throughout the intended coverage area. Typically this equates to at least 3/5 "bars" of coverage as shown on your wireless device, or your device reporting a "good", "very good", or "excellent" signal. This helps ensure that coverage in the intended areas is predicatable and reliable, and is somewhat different to how we have previously deployed wireless (where any signal strength was acceptable).
During the roll out, we are attempting to document the areas of intended coverage (as well as any areas we have explicitly ruled out) on our map of wireless hotspots. Clicking on the marker for a given area will provide this information.
It is likely that in any partial deployment we will cover areas beyond those that are intended (in fact, we deliberately try to do this). Whilst this coverage is useful and desirable, it is also incidental. Thus it is important to understand that even though you may be able to see a wireless signal it doesn't imply that we've explicitly tried to provide coverage in that part of the building. If the signal is weak, then we definately didn't — wireless networks will still work (albeit unreliably) with significantly weaker signals than the minimum we aim for, which means you'll see weak or poor coverage in a much larger area than the intended coverage area.
Incidental coverage, whilst certainly usable, is unpredicatable and therefore cannot be supported. In most instances, the only way to get reliable wireless coverage in these areas will be to install additional access points.
The first phase of the renewal commenced when the site surveys were complete, and was substantively complete by 25 June 2014†. In this phase, about 200 access points were be deployed across campus (a three-fold increase over the current system). Many of these will require new work — additional or moved cabling, etc — and we may outsource some of this installation work.
This phase aims to provide wireless coverage as follows:
- High-profile venues. Three venues — Barratt Lecture Complex, Eden Grove, and the Main Library — will receive comprehensive wireless coverage with a particular emphasis on supporting large numbers (> 500) of concurrent users. All of these are widely used by students and staff alike, and are currently the main venues from which the existing wireless network is used. In addition, both Barratt and Eden Grove are regularly used for high-profile, international, and external conferences & events.
- Like-for-like replacement. The current wireless system will be discontinued, and it will no longer be possible to continue using using the existing wireless access points. Thus, in all areas with existing wireless coverage, this phase will provide like-for-like replacement. In some places, particularly those where comprehensive coverage was the initial goal, coverage may be augmented.
- Large academic spaces. The original wireless deployment aimed to provide coverage of all major lecture venues and seminar spaces that had fixed seating for more than fifty people. However, it has become apparent that a number of venues (for example Geology C11, Psychology Major & Council Chamber) that should have been covered were missed. This will be addressed during the first phase, using the central venue booking system as the main source of information. Should the number of access points allow for it, we may try to cover additional, smaller shared venues.
- Selected transit accomodation. Limited, low-density (< 10 users) wireless coverage will be provided in selected transit houses on campus, particularly those that are frequented by high profile or international visitors. The main aim of this is to make it easier for guests of the University to get access to the Internet during their stay. However, it also serves as a pilot for future coverage in conference accomodation, etc and thus is important from a long-term planning perspective.
- Latent demand. A number of departments and divisions have previously contacted us asking for coverage that we've been unable to provide. Where these requests coincide with our rollout plans (and particularly where the coverage is for multi-use space, or space occupied by students), we'll attempt to address them in the first phase. Before we can make any decisions about placements, however, we'll need to survey the buildings concerned to accurately determine the requirements.
Phase 2 was conducted concomentently with phase 1, starting in December 2013. It was completed at the beginning of June 2014. By 25 June 2014, we had deployed well over 300 access points in over a hundred separate buildings.
During the first half of 2014 we provided wireless coverage in all residence common rooms, funded by the Student Network. This followed the priorities of the telephone project, but was further targeted for completion ahead of the National Arts Festival, where residences will be used to provide guest accomodation.
Whilst the phase one like-for-like replacements were in progress, we tried to anticipate areas that would likely be covered in the next phase of the project and provide coverage for them. This meant that some smaller centrally booked spaces were provided coverage, with the net result that most centrally-booked venues with more than thirty seats (and indeed, some smaller than this) now have wireless coverage. This puts us well ahead of the "large academic spaces" targets of phase one.
In addition, whilst we were undertaking like-for-like replacements, several departments elected to augment the coverage in their building to include areas that, whilst of interest to the departments concerned, did not necessarily coincide with the principles of the renewal project. Realising that there were significant economies of scale in doing concurrent installations, we actively promoted this approach in the later stages of phase 1. Such additional coverage was provided at the deparment concerned's expense.
A previously unaddressed area of thewas that of wardens' telephones. In December 2013, a decision was made to replace existing analogue telephones in wardens' flats with wifi-based SIP handsets. This necessitates the deployment of limited wireless coverage into wardens flats. Budget for this has been made available under the auspices of telephony, and the actual deployment occured alongside student common rooms (see above). This deployment will primarily follow priorities set by the telephony renewal project, although this might be superseded by other requirements (such as conferences accomodated in particular residences).
The Provost and the Library quad (the area surrounded by The Library, English, Geography, GLT, and Geology) were covered out of a budget surplus in the Student Network
The expanding footprint of wireless on campus can be seen in our coverage map.
During the remainder of 2014 we'll likely try to fill in some of the gaps, and cover smaller shared venues, public computer labs, departmental seminar rooms, etc. Our target size for this phase is shared spaces that accomodate thirty or more people, although the budget, ease of installation, and available access points will determine the actual limit. The central space utilisation database will be the primary source of data for determining these areas.
In addition, we will look to those academic departments whose buildings have not already benefited from the first two phases with the intention of providing one access point per department (note, not per building — departments that span buildings will, at least for the moment, be considered covered if at least one building has coverage) in a suitable shared space. This will result in our covering some smaller spaces than those identified in the preceeding paragraph.
Where similar common areas exist in off-campus accommodation, we may consider deploying coverage on request.
Providing comprehensive coverage in residences is expensive and, if it is to be achieved, will take many years. The logistics of doing so are currently unknown. As a trial, during the second half of 2014 we will provide comprehenive coverage in a small number of residences. Calata House was cabled for wireless during construction, and so will be the first beneficiary of this. Thereafter residences will be selected on a weighted lottery system, depending on a commitment from Student Network house reps to facilitate a wireless survey. This is similar to how the cabled portions of the Student Network were deployed, and so the history of that may be informative.
In recognition of their use during conferences and events, and the fact that they are increasingly being used to provide after-hours study space, the nine dining halls will receive coverage during the second half of 2014.
Our ultimate aim is to provide at least some wireless coverage in every reasonably-sized building on campus. Such coverage would typically be focused on "big-win" or high-traffic areas, such as seminar & meeting rooms, laboratories and other teaching spaces, student study space, etc. Priority will be given to areas that are regularly used for inter-departmental/inter-divisional meetings, where people might reasonably be expected to bring mobile devices to meetings. Individual staff offices represent the lowest priority from a central budget perspective, because these areas are already well served by traditional cabled network points, and because the number of devices per access point will typically be very low (making them expensive).
At this stage it isn't clear what budget might be available for future phases, and so it is difficult to predict any timeframes (see below about pre-empting this). However we'll attempt to provide such coverage in as equitable way as possible, aiming to provide wider coverage rather than more comprehensive coverage in a small number of places. Criteria such as the number of rooms and floors in a building, the occupancy, the square meterage and the likelihood of student utilization will all be considered. We will likely take guidance from the Information Technology Division Steering Committee as to institutional priorities.
The first phase of the renewal project was funded from a combination of central budget and Student Network income, on a roughly 50/50 basis.
Thereafter, the rollout split into four distinct streams based on their funding source. Student residences and dining halls were funded by the Student Network; shared spaces like lecture venues will be funded by the IT Division's central budget; wardens' accomodation will be funded by Residential Operations; and some departments/divisions may fund coverage within their own budget. Additional, student-interest areas may be funded out of the Student Network income.
Differing budgets mean that one area might become better served than the other. However, the preceeding section details some of the guiding principles that such future rollouts will be based upon.
Going forward new building projects should include provision for comprehensive wireless coverage. An off-plan site survey will be conducted before building commences to ensure that the optimum locations for access points is known at construction time. Whilst the building project might not provide all the access points needed for such coverage, it will make provision for the cabling requirements so as to avoid the need to retrofit this later.
The new Life Sciences building is the first example of such a building.
Budget permitting, we may consider providing coverage in other areas such as well used outdoor spaces and common gathering points. The area around the Day Kaif is an existing example of such coverage that was augmented during the second phase of this rollout. Some of these areas — particularly those that primarily benefit students — may be funded by the Student Network rather than central budget.
Where a department or division has a specific requirement that will not be met as part of the larger rollout plan, they may request that the IT Division install wireless coverage. However, the costs of such an installation will be for the individual department/division's account and will not come out of the central budget.
This method can also be used to pre-empt the larger roll out plan, and gain wireless coverage substantially sooner than might happen through central sources. A number of departments have suggested they may wish to do this.
To avoid complicating the roll out of the initial phase and to make things more equitable, requests for additional coverage made after 2013-06-04 will not be considered. Once we're in a position to entertain such requests (once the substantive part of the like-for-like replacements is complete†), an announcement will be sent to ict-announce@lists. After the announcement requests for additional coverage should be re-sent, and will be considered on a first-come, first-served basis.
Unlike the system it's replacing, the Ruckus wireless system provides for a wide range of different access points. These range from reasonably cheap, low-end access points that are designed to support small numbers (< 10) of people in a relatively confined space, to high-end access points designed to support large numbers (> 300) of concurrent users. The fully-installed cost of access points varies considerably depending on the model used. We expect that most installations will use the mid-range models (such as the ZF7372). However, depending on the structure of individual buildings and the utilisation of space within them, in some instances it might be preferable to install a single higher-density access point rather than several lower-density ones; in others the situation might be reversed. The costs also vary with the USD exchange rate. Our price list reflects the current costs of access points.
It is our intention that all new wireless deployments form part of a long-term plan for comprehensive coverage. Rather than simply making ad-hoc deployments on demand, we will site survey new locations to determine the optimum locations for access points. This will allow us to incrementally grow coverage as required, and avoid (expensive) relocations in future.
Thus rather than having a fixed cost for wireless installations, we'll make a recommendation based on the actual requirements and the structure of the building.
It would then be up to individual departments/divisions to source the requisite funds from their own budgets.
† It was intended that this work be completed before the end of 2013. However a number of issues, including an injury-on-duty of one of the staff members involved in deploying wireless, put this significantly behind scheduled.
Last Modified: Fri, 17 Feb 2017 14:26:43 SAST