Scaling a mobile tutoring project: strategic interventions in C3TO

Scalability is a critical characteristic when attempting to take a successful research  "pilot project" and roll it out on a large scale.  Dr Math, an online mobile tutoring system,  was such a pilot project. Dr Math linked primary and secondary school pupils to university students using Mxit on the pupils' cell phones in order to provide the pupils with quality help with their mathematics homework in the afternoons after school.  From an initial expectation of approximately 50 pupils, Dr Math grew to assist thousands of pupils during its initial 3 years of existence.  However, as more and more pupils asked Dr Math for help with their mathematics homework, it became clear that the original IT architecture could not scale to handle the unexpected work load.  C3TO, Chatter Call Centre/Tutoring Online, specifically addresses the scalability issues of online mobile tutoring. Scalability is the ability of an IT solution to satisfactorily handle work loads of different sizes.  C3TO tackles the scalability problem at three different levels:  technical, tactical, and strategic.  The technical level includes "what" components are being used in the solution (for example, Mobicents running under  Jboss on Linux servers).  The tactical level includes "how" these components are configured, accessed, and used. The strategic level includes facilities which attempt to answer the pupils questions using the least amount of resources and protects scarce resources from misuse (or abuse).  Dr Math was ported to the new C3TO architecutre in November, 2009.  At the time of writing this abstract, a handful of tutors from various tertiary institutions around South Africa, are using C3TO to provide quality mathematics tutoring to hundreds of pupils.  This paper will discuss in detail the technical, tactical, and strategic solutions required to enable a mobile tutoring system to scale adequately.  In addition, in view of the fact that this conference is to be held later in 2010, we will have more numerical data on scalability issues such as tutor/pupil ratio, connectivity, data requirements, etc.