Maaz Mohiuddin’s research while affiliated with Swiss Federal Institute of Technology in Lausanne and other places

We consider the problem of coordination among replicated SDN controllers, where the challenge is to ensure a consistent view of the network while reacting to network events in a prompt manner. Existing solutions are either consensus-based, which achieve consistency at the expense of high latency; or eventual-consistency-based, which have low latency at the expense of severe limitations on the types of applications and policies implementable by the controller. We propose the Fast and Consistent Controller-Replication (FCR) scheme. FCR is based on a deterministic agreement mechanism that performs agreement on the input of controllers, instead of agreement on the output as done in consensus mechanisms. We formally prove that FCR provides the same guarantees in terms of implementable applications and network policies, as any deterministic single-image controller. Through simulation and implementation, we show that these guarantees can be implemented with little latency overhead, compared to eventual-consistency approaches, and can be achieved significantly faster than consensus-based approaches.

Real-time control systems (RTCSs) perform complex control and require low response times. They typically use third-party software libraries and are deployed on generic hardware, which suffer from delay faults that can cause serious damage. To improve availability and latency, the controllers in RTCSs are replicated on physical nodes. As physical replication is expensive, we study the alternative of exploiting virtualization technology to run multiple virtual replicas on the same physical node. As virtual replicas share the same resources, the delay faults they experience might be correlated, which would make such a replication method unsuitable. We conduct several experiments with an RTCS for electric grids, with multiple virtual replicas of its controller. We find that although the delay of a virtual machine is higher than of a physical machine, the correlation between high delays among the virtual replicas is insignificant, causing an overall improved availability. We conclude that virtual replication is indeed applicable to certain RTCSs, as it can improve reliability without added cost.

In real-time control of electric grids using multiple software agents, the control performance depends on (1) the proper functioning of the software agents, i.e., absence of software faults, and (2) the behavior of software agents in the presence of non-ideal communication networks such as message losses and delays. To evaluate the control performance of such systems, we propose T-RECS, a virtual commissioning tool. T-RECS enables testing the performance of software-based control in-silico (before the actual deployment of software agents in the grid), saving both time and money. Developers can run the binaries of their software agents in T-RECS where these binaries exchange real messages by using an emulated network and simulated models of the electric grid and resources. Consequently, the control of an entire microgrid can be tested on a standard computer. In this paper, we first describe the design and the open-source implementation of T-RECS. Second, we measure its CPU and memory usage and show that our implementation can accommodate eight software agents on a standard laptop computer. Third, we validate the simulated grid used in T-RECS by replaying data collected from experiments performed in a real low-voltage microgrid. We find that the average error is 0.037% and the 99th percentile of the error is less than 0.1%. Finally, we present some typical use-cases of T-RECS such as performance evaluation (1) under extreme grid conditions and (2) with non-ideal communication networks. The former, i.e., performance evaluation under extreme grid conditions, is difficult to test in the field due to safety concerns.

We compute bounds on end-to-end worst-case latency and on nodal backlog size for a per-class deterministic network that implements Credit Based Shaper (CBS) and Asynchronous Traffic Shaping (ATS), as proposed by the Time-Sensitive Networking (TSN) standardization group. ATS is an implementation of the Interleaved Regulator, which reshapes traffic in the network before admitting it into a CBS buffer, thus avoiding burstiness cascades. Due to the interleaved regulator, traffic is reshaped at every switch, which allows for the computation of explicit delay and backlog bounds. Furthermore, we obtain a novel, tight per-flow bound for the response time of CBS, when the input is regulated, which is smaller than existing network calculus bounds. We also compute a per-flow bound on the response time of the interleaved regulator. Based on all the above results, we compute bounds on the per-class backlogs. Then, we use the newly computed delay bounds along with recent results on interleaved regulators from literature to derive tight end-to-end latency bounds and show that these are less than the sums of per-switch delay bounds.

Real-time control systems use controllers that compute and issue setpoints within stringent delay constraints. Failure to do so, due to a crash or delay as a result of software and/or hardware faults, can cause failure of the controlled resources. Recently, Axo, a protocol for masking crash and delay faults by replicating the controller, was proposed. Axo provides safety by discarding delayed setpoints, and it relies on the presence of valid setpoints for providing availability. To ensure that enough valid setpoints are issued, faulty controller replicas need to be detected and recovered. We present a mechanism for detection and recovery of delay- and crash-faulty replicas under the Axo framework. These mechanisms were designed to be soft state (i.e., their state can be reconstructed from received messages) to enable seamless additions of new replicas. Besides presenting the design, we analytically characterize the time to detect and recover a faulty replica, and we validate them experimentally. We demonstrate the performance of Axo by using two case studies: the first provides a stability analysis of an inverted pendulum system with Axo, and the second shows the fault-tolerance performance of Axo through a deployment on a real-time control system that controls a CIGR'E low-voltage benchmark microgrid.

Applications performing streaming of phasor-measurement data require low latency and losses from the communication network. Traditionally, such requirements are realized through wired infrastructure. Recently, wireless infrastructure has gained attention due to its low-cost and ease of deployment, but its poor quality-of-service is a strong deterrent for use in mission-critical applications. Recent studies have used measurements to explore the use of packet replication over redundant Wi-Fi paths, for obtaining the desired loss performance without hampering the end-to-end latency. However, these studies are done in a controlled, laboratory environment and do not reflect the real, in-field performance. In this paper, we perform extensive measurements using two co-located directional Wi-Fi links in a real-life setting, to experimentally validate the use of packet replication over Wi-Fi for streaming phasor data. In the setting that we evaluated, we find that the two channels are not fail-independent but the performance achieved with replication is very close to what it would be if they were to be independent. From the loss and latency statistics after replication, we conclude that replicating the phasor data over redundant Wi-Fi paths is a viable option for achieving the desired quality-of-service.