黑料福利网

Ensuring that technologies consistently meet their timing requirements is becoming increasingly difficult as systems grow larger, more connected, and more complex. In her PhD research, developed new methods that help engineers analyse and verify the timing behaviour of these systems before they are deployed. She defended her PhD thesis at the Department of Mathematics and Computer Science at 黑料福利网 on Wednesday, July 1.

Predictability in an uncertain world

Many modern technologies belong to a class of systems known as real-time systems. These are systems in which producing the correct result is not enough; that result must also be delivered within strict timing limits.

The challenge is that such systems rarely behave in exactly the same way twice. Tasks may start at slightly different moments, require varying amounts of time to complete, depend on the completion of other tasks, or temporarily pause while waiting for information from elsewhere in the system.

Consider a modern factory, where multiple machines and software components continuously coordinate their activities. One task may have to wait for another to finish before it can start. A machine may need information from a sensor before continuing its operation. Small variations like these can create a huge number of possible execution scenarios.

As systems become increasingly interconnected, this uncertainty grows rapidly. Yet engineers still need guarantees that critical tasks and messages will always meet their deadlines.

More realistic system behaviour

Modern systems do not simply perform computations. They constantly switch between processing information, waiting for data, communicating with other devices, and resuming their work. Understanding the timing of all these interactions is essential for ensuring reliable behaviour.

Many existing timing-analysis techniques rely on simplified assumptions about how tasks behave. In practice, however, real-world applications are far more complex.

Tasks may consist of multiple stages, depend on previous activities, or temporarily suspend themselves while waiting for information from sensors, remote computers, or other components within the system. During these waiting periods, other activities continue to execute, making the overall timing behaviour much harder to predict.

Srinivasan developed new methods that explicitly account for these realistic forms of task behaviour. By modelling dependencies between activities and periods of waiting, her research enables engineers to analyse systems that more closely resemble those used in practice.

Keeping communication on schedule

Communication is becoming just as important as computation. Modern real-time systems increasingly rely on networks to exchange information between devices, meaning that not only computations but also messages must arrive on time.

Consider a factory where sensors continuously monitor a production process, controllers make decisions based on that information, and robots carry out physical actions. These components often operate on separate computers connected through a network. If a control message arrives too late, a robot may act on outdated information or miss a critical deadline altogether.

Similar challenges arise in automotive systems and other distributed technologies where multiple devices must coordinate their actions in real time.

As part of her research, Srinivasan investigated timing guarantees in Time-Sensitive Networking (TSN), a technology designed for communication networks that require predictable timing behaviour. Her methods help engineers determine how long messages may take to travel through a network and whether timing requirements can still be guaranteed under different operating conditions.

Finding order in complexity

One of the biggest challenges in timing analysis is the enormous number of possible ways a system can behave.

Checking every possible scenario individually quickly becomes impractical. As the number of tasks, interactions, and sources of uncertainty increases, the number of possible execution paths can grow dramatically.

To address this challenge, Srinivasan built upon an existing framework known as the Schedule Abstraction Graph (SAG). Rather than analysing every possible system behaviour separately, the framework identifies scenarios that are effectively similar from a timing perspective and analyses them together.

This can be compared to studying traffic patterns in a city. Instead of examining every individual journey one by one, similar routes can be grouped together to understand the overall flow more efficiently. By applying a similar principle to timing analysis, Srinivasan developed techniques that significantly reduce the number of scenarios that need to be explored while still ensuring that no important behaviour is overlooked.

This makes it possible to analyse larger and more complex systems while maintaining confidence in the results.