Twenty delegates from twelve different organisations representing operators, vendors, consultants and academics attended the 16th annual STEM User Group Meeting from 05–06 October 2011.
The two-day event at King’s College in Cambridge, UK combined two complementary themes:
- a
modelling track, including a showcase technique for modelling upgrades in a mobile network,
plus training sessions on geographical overheads for peak-traffic contention and required
equipment capacities
- a technology track more specifically focused around developments, both current and future,
in the STEM modelling platform itself.
The full proceedings of the event are now available as a PDF download.
ACG Research |
Ciena |
Queen Mary, University of London |
Aetha Consulting |
Cisco |
Sky Network Services |
Analysys Mason |
Investaura |
Telkom SA |
Chemnitz University of Technology |
Juniper |
Welsh Government |
Figure 1: Organisations represented at the STEM User Group Meeting in 2011
Modelling track with an evident statistical theme
The first session queried how you would design an efficient, macro-economic simulation
of an irregular series of capacity upgrade steps deployed
separately at each base-station site in a mobile network where the local growth
in demand (and timing of required upgrades) varies by site? We then unveiled a readily
parameterised STEM model which makes a realistic and justifiable simulation
of the actual requirement without having to ‘hire a graphics card’ to
do all the processing or an ‘army of helpers’ to gather countless inputs
per site.
A later training session used STEM to dimension a school bus fleet for a local authority
and involved audience participation to generate a
random distribution of pupils requiring lifts per school.
The amateur statistical theme returned at the start of day two with a further training
session examining rush-hour planning for a taxi firm
as a proxy for the kind of calculations ISPs do to allow for everyone switching
on the iPlayer at the same moment for the BBC News at Ten.
A planned session on ring dimensioning (as previewed
in our July newsletter), with the interesting contrast that certain groups of calculations
should not vary continuously by site, was skipped by popular consent to allow time
for a series of four guest presentations arranged in true consultant style at the
eleventh hour!
Guest presentations
Thomas Knoll (Chemnitz University of Technology) gave a talk on the techno-economic challenges within the LTE project MEVICO and the use of STEM in it.
Manfred Illenberger (Cisco) presented a neural-network approach for reducing model run-times – especially relevant in a web context – conceptually equivalent to compressing or ‘zipping’ the calculation logic, irreversibly replacing the model structure with a ‘hard-wired’, much simplified and faster core with limited inputs and outputs.
Steve Liput (Juniper) delivered a refreshing tutorial on the use of the Function element in STEM to group equivalent resources and implement automated handover or technology update from one obsolete resource to its future replacement.
Pierre Lurin (Investaura) described how STEM’s built-in cost allocation capability provides insight into the unit and incremental costs of service provision, with drill-down into exactly which network and process costs are influenced by each service. A model built for a large mobile operator is used to analyse the past, present and future profitability of individual services, tariff plans and customer segments, and provides results to optimise current tariffs and serve as inputs to the pricing team for new voice and data services.
Figure 2: Intimate audience involvement during the STEM User Group Meeting sessions
Technology track
In session two, we staged a live demonstration of putting a STEM model on the web from scratch, getting the audience to choose inputs for the online interface to prove we weren’t cheating. The process worked flawlessly, if a little slower than planned: our VPN connection was unexpectedly routed over very contended mobile ‘broadband’ (rarely exceeding 1 kbps!) due to unforeseen limits in the venue’s WiFi provision. Nevertheless, delegates were able to access the model over the public Internet from their own WiFi-enabled devices within the 90 minutes scheduled.
After the excitement of the after-lunch training interlude, and with apologies in advance for the high level of technical content, we walked through a huge list of historical bugbears now tamed in the forthcoming STEM 7.3.
In a presentation on the second morning, which could be described as ‘heavy’, we set the record straight about how incremental quantities will interact with shorter time periods in STEM 7.3 (including an impromptu model of birthday-cake consumption) and did little to settle the protracted debate about suitable terminology for these new concepts.
In the final session, we took the usual opportunity to talk more generally about future development directions, both online and on the desktop (or in your pocket). Unlike similar sessions in past years, it quickly became evident that both the audience and Implied Logic had plenty of ideas to air! The general consensus was that hosted model access with eSTEM is only the first step towards a future in which users may collaborate on STEM model development in a server context.
Please register to access the proceedings
When you click through to access the proceedings in PDF format, you will be required
to complete a one-time registration process which will register your preferred email
address as an account log-in for this and any future access to restricted areas
of our website, with an opt-in check-box for our quarterly email newsletter (or
opt-out for existing recipients who no longer wish to be included).