Provisioning, Enforcing, and Pricing Temporal Service Differentiation in Virtualized Networked Environments. This is a collaborative project with Profs. Chris Gill and Chenyang Lu at Washington University. The project is supported by NSF CNS-1514254, and its main goas are to improve the performance of the network stack in virtualized systems, using Xen as its main experimental platform. The motivation is that while OS level schedulers are capable of efficiently partitioning compute resources across virtual machines (VMs), and network mechanisms are similarly efficient in enforcing appropriate bandwidth partitioning between competing traffic flows, the same does not always hold when dealing with the network stack responsible for arbitrating network access between VMs sharing the same network connection.
The project, therefore, seeks to develop new approaches capable of offering guarantees to competing VMs with different performance requirements. Of particular interest is differentiating between VMs with low latency requirements versus VMs with high throughput goals. Ultimately, the project seeks to develop a comprehensive understanding of the trade-off between allocating additional CPU resources to managing access to network resources between competing VMs, and the strength of the service differentiation capabilities it affords. The broader goal is to develop and demonstrate mechanisms that are both effective in enforcing service differentiation and frugal in their own resource requirements. In particular, this means that solutions should not call for dedicated hardware support.
PaaS Message Communication Middleware. This is a collaborative project with Profs. Chris Gill and Chenyang Lu at Washington University. The project is supported by Huawei, and explores challenges and solutions for delivering a lightweight messaging middleware capable of delivering latency differentiation.
Protocol Stacks Design and Evolution: The Role of Layering and Modularity. This is a collaborative project with Prof. Constantine Dovrolis at Georgia Tech. The project is supported by NSF grant CNS-1319684, and its main goal is to shed some light on factors that contribute to the success and evolvability of network protocols.
The project initially focuses on the development and analysis of several models that seek to on one hand capture key aspects/features of protocol design, and on the other hand allow the systematic investigation of how differences in those features affect a protocol’s success and in particular its adoption and ability to spawn a growing usage. A subsequent step in the investigation will be to translate the insight derived from those findings into either protocol instances that would inherit those properties, or into possible suggestions for modifications to existing protocols to improve their ability to evolve.
J. Liu and R. Guerin, “Multipath and Rate Stability.” Proc. IEEE Globecom 2016 - CQRM: Communication QoS, Reliability & Modeling Symposium, Washington, D.C., December 2016.