'Sieve' is a platform which derives actionable insights from monitored metrics in microservice applications. An example of 'actionable insight' is a set of rules for auto-scaling of microservices in Cloud deployments. On top of it, I have worked on a non-intrusive Root Cause Analysis method for anomalies detected in microservice applications, tested over a containerized deployment of Openstack.
Work developed while interning at NOKIA Bell Labs, in Stuttgart, Germany, under the supervision of Ruichuan Chen and Istemi Ekin Akkus. Plus, I came up with the system's name :P.
Paper in Middleware'17
View on GitHub
Studied the tradeoffs involved in the usage of different types of content identifiers used in Information Centric Networks, namely hierarchical names and flat IDs. Focusing on the property of aggregation - essential for routing scalability, and exclusive to hierarchical identifiers - I’ve designed and implemented a method which allows for the usage of hierarchical names over flat ID ICNs. This method consists in encoding the inner prefixes of names into Bloom filters - which are effectively fixed-sized bit strings - keeping the property of aggregation.
I have also studied the influence of false positive matches on forwarding correctness. The study shows that forwarding errors increase when aggregation levels are high, uncovering a trade-off between forwarding efficiency and routing scalability. I also demonstrate the benefits of ‘namespace’ planning and in-network error resolution to mitigate the negative impact of such errors.
Paper in ICN'18
Data used in ICN'18 paper
Extended version of ICN'18 paper
False Positive Analysis tool, using Rocketfuel topologies (used in ICN'18 paper)
Implementation over the eXtensible Internet Architecture