– Navneet Sharma, COO, ACT

An online examination can be disrupted in a matter of minutes. A portal may not load, a question paper may fail to download, a proctoring connection may drop, or a student may lose valuable time because the digital infrastructure cannot keep up with demand. What appears to be a technical issue can quickly become an issue of fairness when some students are able to continue while others are forced to wait.

Recent examination disruptions show that this is not a hypothetical concern. In May 2026, a technical glitch affected the CUET-UG examination at several centres. The National Testing Agency (NTA) subsequently announced a re-test for 3,765 candidates who had completed biometric registration but left the examination centres before the test could resume. In 2025, NTA also cancelled a CUET-UG examination shift at a Srinagar centre because of technical glitches, affecting 76 candidates who were subsequently scheduled for another examination. These incidents highlight the importance of designing academic networks and digital examination infrastructure for peak examination conditions rather than average campus usage.

Coverage, capacity, continuity and control must work together

Bandwidth at the internet gateway is only one part of the challenge. A campus can have a high-capacity connection and still experience poor connectivity in a particular classroom, laboratory or hostel. Institutions therefore need to assess capacity at the level where students actually connect, which means access point density in every examination hall. Coverage on a heat map is not the same as capacity per device, and a hall that comfortably supports 40 students in a lecture behaves very differently when 300 devices authenticate within the same five minutes.

Dynamic load balancing addresses this. Intelligent access points can detect when a node is approaching its client limit and steer devices to adjacent, underutilised access points, preventing a single-point connection crash in one hall while the rest of the network reports healthy utilisation. Capacity planning should also reflect the shape of examination demand, which is short, sharp and concurrent rather than steady. Burstable bandwidth, which ACT offers on its enterprise connectivity, allows an institution to draw above its committed capacity during a defined examination window without paying for peak capacity year-round.

Control: when the network is healthy but clogged

In high-density campus environments, the physical access points often hold up. The failure comes from elsewhere. Cloud syncs, operating system updates, streaming and social media running quietly on students’ devices saturate the same airtime and uplink the examination portal depends on. The network is physically healthy and functionally clogged.

Traffic control, or Quality of Service, is therefore the fourth requirement. Peak bandwidth should be reserved explicitly for examination domains, proctoring video streams and biometric authentication APIs, with guaranteed queues rather than best-effort treatment. Temporary SSIDs or VLAN isolation, supported by ACT as part of its managed enterprise connectivity, keep assessment traffic on a separate logical network so an incident elsewhere on campus does not degrade an examination in progress. Non-essential traffic categories can then be throttled for the duration of the window and restored automatically afterwards, applied as a scheduled policy rather than a manual intervention someone has to remember to reverse.

Continuity that protects the session

For an examination, last-mile redundancy means having an alternate path available if the primary connection fails. The mechanics matter. Dual dedicated internet leased lines from independent providers, in an active-active configuration with automatic failover, allow traffic to continue over the surviving path without a manual switch. ACT provides multi-ISP auto-failover on its enterprise leased lines for this purpose. The objective is not only that the campus returns online quickly, but that the session itself does not disconnect, time out or reset the candidate’s saved state.

Proctoring adds a requirement that raw speed does not address. AI-based proctoring and live invigilation depend on low latency and stable jitter far more than on download throughput. A link that tests well but fluctuates under load can still produce frozen video, failed identity checks and false flags against honest candidates.

Testing before students do

Centralised monitoring and network analytics help institutions understand where devices are concentrated and which access points are approaching capacity. What reveals genuine weakness, though, is a simulated concurrency test that reproduces the real examination profile: several thousand devices authenticating within minutes, downloading a question paper of realistic size at the same moment, then sustaining proctoring streams for the full duration. A steady-state test on a quiet Sunday proves very little.

A practical pre-examination sequence covers peak concurrency hall by hall, weak coverage and high-density zones, end-to-end testing of authentication and examination platforms, confirmation that critical applications hold their priority, and validation that failover works while the network is under load rather than idle.

Towards an exam-day IT standard

Much of this is currently treated as network maintenance. It deserves to be treated as institutional policy. An examination-day IT standard, adopted formally and reviewed each cycle, can define the minimum access point density for an approved hall, the traffic priority rules that apply during a live window, the failover expectation stated in seconds, the testing that must be signed off before a hall is cleared for use, and who decides within how many minutes when an examination is disrupted. Written down, these become commitments for which an examination controller and an IT head are jointly accountable.

Building a network that protects examination fairness

When one student loses time because a question paper will not load or an access point becomes overloaded, the impact is not limited to the IT department. It affects that student’s ability to perform in a high-stakes assessment.

The question should therefore move beyond, “How much bandwidth do we have?” to, “Can our network continue to support every student when thousands of users need critical examination services at the same time?” As higher education becomes increasingly digital, reliable connectivity should be treated as part of examination readiness. The real meaning of bandwidth equality is not that every application receives the same amount of bandwidth. It is that every student gets a fair opportunity to connect, participate and complete the examination without the network standing in the way.

Also Read: Teaching Design in the Age of Generative AI