A primer for learner-centric bandwidth capacity planning
Managing bandwidth requirements for educational institutions is a challenging and complex task, particularly for IT Directors caught between increased expectations of ICT service delivery and diminishing budgets.
Digital transformation in the education sector has presented a new set of challenges around meeting the bandwidth expectations of individual learners.
To manage these challenges, AARNet’s Nick Cross suggests educational institutions – primary, secondary and tertiary – take the approach of correlating current and future capacity planning to individual learning opportunities. To help IT Directors with this task, here he offers a logical structured approach for determining requirements across the end-to-end digital supply chain.
Building a Hierarchy of Needs
In the seminal paper ‘Understanding Students’ Uses of technology for Learning – Towards Creative Appropriation’, Sharpe & Beetham (2010) suggest there are a range of pre-conditions that must exist for a learner to flourish in a technology-rich learning environment.
Adapting the form and function of Maslow’s Hierarchy from ‘A Theory of Human Motivation’ (1943), Sharpe & Beetham offer a developmental model of learning in a digital context.
At the base of the pyramid is the requirement to access technologies, resources and services. This forms the bedrock upon which all other digital capability is facilitated. Without reliable, convenient and effective access, none of the other attributes of effective [self-directed] learners can be bought into play.
The functional access tier incorporates a range of elements that is inclusive of access devices, security and access policy, governance, trust, the built environment and, for the purposes of this paper, the underlying network infrastructure. Note that the addition and emphasis of the word functional is to be interpreted from the context of the learner.
Conceptualise your Digital Supply Chain
A network can be thought of as an interconnected or interrelated chain, group, or system bridging assets or value, the components of which from a functional perspective form an end-to-end digital supply chain. In the learning context the digital supply chain may have at its western node an inquiring mind, poised to satisfy some curiosity. At its eastern node is the resource or entity that can contribute to the appeasement of that curiosity. In between lie the network elements that are called to autonomously coalesce across disparate administrative domains to honour the ‘What if?’ moment. At the learner level this is a one to many relationship.
I recall a recent conversation with a University CIO who stated, ‘we meet them [students] on their device.’ This is both a practical and insightful allegory of delivering a user-centric service model in a manner that is on the learner’s terms. It is also a very personal statement and a commitment of his institution’s obligation to construct an entry point for discovery as defined by the learner.
The challenge at an institutional level is that there are hundreds or thousands of these moments of inquiry simultaneously procreating at a single point of time, each of which places a service expectation on the network. At the institutional level this is a many to many relationship.
What is learner asking of the network? What are the characteristics of the associated packet flows and how can this then be quantified to inform network capacity planning to reliably meet this demand on a per learner basis?
Understanding application-centric dimensioning
A useful starting point is to target two fundamental metrics – throughput and latency. Real time applications such as Unified Communications have bought a focus to network dimensioning and responsiveness as a pre-requisite for reliable service delivery. This same utility can be applied to all applications traversing the network and beyond a historical best effort mentality.
For general web browsing and editing Google Drive documents, 0.2-0.5 Mbps per concurrent session at a minimum should provide satisfactory performance. If students will be streaming video or using video collaboration, at least 1 Mbps per concurrent user session is needed and >4 Mbps is required for HD video streaming.
Informed by testing and/or vendor guidance, a matrix can be developed which provides a profile of the applications used by your on-campus community and to ascribe appropriate throughput and latency metrics on a per session basis. Firewall/Proxy logs can often provide a list of applications at use on the network and commonly offer other useful metrics, including number of active sessions, and total and average volumes per session.
When testing for latency it is important to identify an appropriate Latency Validation Target (LVT) relevant to the application being tested. This is essential for critical and rich-media applications as the manner by which packets traverse the carriage domain may be substantially different on a per application or component of application basis. From this point it is possible to characterise application requirements and extrapolate based on the size of the institution and modelled on ‘number of active concurrent session’ scenarios.
Table 2 is an example scenario based on an institution of 1,000 students, conservatively modelled on a 10% concurrent session ratio across differing application classes that provides a range of estimates of the aggregated network capacity required. This will vary for each institution and can be applied according to a local context.
Table 3 provides a conservative modelling for an institution of 1,800 students modelled on an 8% concurrent session ratio.
For many institutions the throughput and latency characteristics of the of the carriage domain are the primary determinant of the end-user experience, which is the cardinal metric by which the efficiency of an infrastructure initiative is reliably measured.
When assessing the service offering of a provider in the carriage domain, it is important have them explain how they have architected their network to refine access to the resource domain ie: do they have direct peers with major content providers (Google, Microsoft, Apple, Akamai) and at what capacity? What is the network cost in terms of distance (hops) and latency to reach these resources? These items are key considerations in optimising this area of the digital supply chain and can impact significantly on end to end latency targets.
It is also important to consider latency and throughput together as they are inter-related. Latency is typically the first casualty of war when network services are oversubscribed and transition to a state of contention. Throughput metrics alone are often the only metric examined when assessing network performance but this approach provides an astigmatic insight to network serviceability with latency being the more pertinent of the two indicators of end-user experience.
Each year AARNet experiences an on average 50% increase in demand on the network. This is motivated by the increase in the range and performance characteristics of offerings in the resource domain, each with escalating per session demands. Whilst not in a purely educational context, the introduction and uptake of Netflix and other on-demand video services are a good example of services that have rapidly escalated demands of the network.
Capacity Management and Planning
Adopting an application centric, metric based approach to network dimensioning offers a strategic pathway to establish the requisite functional access foundation.
With cloud based service delivery for core services (apps, docs, LMS, messaging, information systems) now commonplace, establishing a baseline of per user access across the digital supply chain is a mandatory pre-qualifying step to expanding the institutional service portfolio. This approach can also reliably inform forecasting for future initiatives and can be included as a baseline target in strategic planning documentation.
For the IT Director caught between increased expectations and diminishing budgets this approach is a useful vehicle for more directly correlating network provisioning to the individual learning opportunity, and an excellent tool for making the business case for appropriate investment in what is now a core competency for an educational institution.
This can also be represented in the form of a student SLA (Service Level Agreement), a service assurance that for some high percentage of the academic day, each student will have at their bidding n Mbps of bandwidth with an associated application appropriate latency profile. A student SLA is a tangible representation of an institutions commitment to each individual learner and an acknowledgement that the functional access tier is an essential foundation to their academic progression within that institution.
Questions about this content or for help with capacity planning, please contact us