The Tesseract is a distributed graph processing and computation platform, purpose built, from the ground up.
This talk will present the new algorithms developed to create an efficient platform for native distributed graph traversals and computations.
The last 10 years has had provable uses of graph processing systems. As we aim to do more, we stretch existing technologies far beyond what they were intended for. Remarkable efforts like Apache Giraph or the Titan project are examples that are pushing the bounds. Ultimately the brute force techniques are inefficient, often don't scale very well and come with their own problems. In the ideal world, a scalable graph processing system would evenly distribute both vertices and edges. Richard Karp has shown this to be an np-complete problem (i.e. the clique problem / complete subgraphs).
The talk will give a brief introduction to CRDT - Convergent & Commutative Replicated Data types and how commutative & idempotent operations enable the three contributions.
Three key contributions The Tesseract makes include partitioning "super vertices" with a technique called "cascading vertices", an optimization called "wormhole traversal" and a computational framework, "TQL" able to make use of two aforementioned techniques. The talk will give an introduction to the first two and a very brief overview of the third.