[Nets-seminars] Seminar today Friday 12/11/10 GS/302 16:00

[Richard G. Clegg]

Don't forget tomorrow's seminar.  Dr Eva Kalyvianaki will be visiting 
from Imperial.

Title: Data Streaming Query Planning

Abstract:
When users submit new queries to a distributed stream processing system
(DSPS), a query planner must allocate physical resources, such as CPU
cores, memory, and network bandwidth, from a set of hosts to queries.
In this talk I will describe query planning in two different settings.

First, I will talk about query planning that targets DSPSs where
allocation decisions must provide the correct mix of resources required 
by queries for their processing. I will describe SQPR, a query planner
which exploits overlap between queries and reuses partial results to
save resources and increase the number of admitted queries. SQPR models
query admission, allocation, and reuse as a single constrained
optimisation problem and solves an approximate version to achieve
scalability. Our experimental evaluation in comparison with a
state-of-the-art planer shows that SQPR makes efficient resource
allocation decisions, even with a high utilisation of resources, with
acceptable overheads.

Second, I will discuss query planning that allocates resources fairly
between queries in a resource-constrained environment. As new queries
are added in a DSPS, processing nodes may overload causing queries to
compete for limited resources. I will describe, FISSP, a query planning
framework which accepts new queries in overload conditions, by
migrating resources from existing queries to new ones through controlled
degradation of processing quality. By balancing resources between
queries, it achieves processing fairness in a distributed manner.


Short Bio:
Eva Kalyvianaki is a Post-Doctoral Researcher at the Department
of Computing, Imperial College London. Before this, she obtained
her PhD from the Computer Lab, University of Cambridge. Her main 
research interests are in managing distributed systems and autonomic 
computing.  She is currently working on building dependable 
Internet-scale data streaming applications. http://www.doc.ic.ac.uk/~ekalyv


-- 
Richard G. Clegg,
Dept of Elec. Eng.,
University College London
http://www.richardclegg.org/