Our Projects

NGLA: Next Generation Log Analytics

NGLA.pngComputer systems generate a huge amount of heterogeneous logs. Those logs provide rich contextual information describing system status and are critical sources for system monitoring and diagnosis. However, manually interpreting those logs is not effective due to the extremely large volume and complicated domain-specific syntax and semantic knowledge. NGLA is a comprehensive and scalable framework to analyze heterogeneous logs from any source without prior domain knowledge or pattern information. It provides a self-learning engine and a stream processing platform for new applications including system anomaly detection with deep log inspection and unstructured log management.

CloudVue: Multi-tenant Cloud Management Platform

CloudVue.pngMulti-tenant data centers use network virtualization to offer each customer the illusion of a personal, dedicated network infrastructure. However, virtualization leads to increased management complexity as network operators have to juggle hundreds of virtual networks in addition to managing the original physical infrastructure. CloudVue improves and streamlines virtual network management by collecting, organizing, and analyzing all performance and configuration information from all layers and devices in the network. It then uses data analytics algorithms to extract and correlate information from various sources and trigger warnings when network performance degrades.

Clue: System Debugging with Deep Analytics

clue.pngModern computer systems, from single servers to large cloud deployments, generate billions of events that reflect the state and operation of the system. CLUE provides a black-box, unsupervised debugging tool to mine event patterns and diagnose performance issues in these systems. CLUE uses novel data mining technologies for automated information retrieval and a state-of-the-art debugging toolset to integrate and profile event transactions.

SDN Monitoring with Zero Measurement Cost

openflowlogo.pngSoftware-defined networking (SDN) enables abstractions that separate the control and forwarding functions in network devices. We study the implications and applications of SDN in network monitoring and management by developing novel traffic measurement solutions that use the power of SDN to monitor the network at little to no cost.