Postdocs, Graduate Students, and Undergraduate Students
Challenges to Scaling Cloud Computing: Gigantic power consumption and Grid transformation to volatile Renewable Generation
Cloud computing is the fastest growing consumer of
electric power in the world! This is a challenge for
the continued growth of the Hyperscalers (Google, FB,
Microsft, Amazon, Alibaba, Baidu, Apple, etc.). We study opportunities
to make distributed systems more flexibly reliable -- allowing them to exploit "Opportunity power" or "Stranded power" produced by power markets, and rising levels of volatile renewable generation.
Topics include distributed systems, resource management, new volatile
services models, serverless, power markets, power grids.
Computing the Representation to optimize Data Movement and Storage
Its widely recognized that data movement (from memory,
from SSD, within a parallel machine, even in the wires
across a chip) is the critical cost and performane
limiter in computer systems. We are studying architectures
that enable efficient rapid transformation of information encodings,
to reduce size and computation cost. Initial designs and studies show that
benefits of 4x to 1000x can be achieved in specific cases.
Critical challenges include how to expose these new ideas
to software: (e.g. transformer libraries, or to view C++ arrays as abstract data types with a
different concrete type implementation), as well as a variety of functional and
implementation architecture issues. The space for this radical changes
has just begun to be explored. We are working with
several agencies to build large scale prototypes (1000's of nodes) within
the next 3-5 years.
Data-intensive computing, Scalable In-memory analytics In-situ Data Analytics
Application of efficient data transformation accelerators
to produce breakthrough data analytics performance (10-100x
absolute performance and power efficiency). Topics include
filtering, RAW processing, efficient ETL, predicate
push-down, and query optimization. Current work is in
SparkSQL with at-scale experiments on a variety of novel