IEEE Technical Seminar by Prof. Olgica Milenkovic on Distributed Storage
Title:
Distributed storage systems with dynamic access/load balancing
Abstract:
The problem of designing codes for distributed storage systems has received significant attention due to its relevance to many practical applications and interesting analytical challenges.
Many relevant distributed coding schemes are based on combinatorial designs which ensure efficient content regeneration and system recovery. An example includes fractional repetition codes that are both conceptually simple and easy to implement.
One important problem mostly overlooked in coding for distributed storage pertains to access balancing (i.e., load balancing) where one requires that the number of access requests to servers both during repair and data download are close-to-uniform. To implement such a balancing scheme, one needs to take into account the popularities of the stored data chunks or files as these dictate the number of access requests received by the servers.
We introduce new access (load) balancing fractional repetition codes that represent the first instance of combinatorial designs with both classical intersection and labeling (coloring) constraints. The codes maintain the properties of classical fractional repetition codes but also ensure that the average popularity scores of data files at different servers are nearly the same. We then describe a dynamic popularity-change model and explain how it relates to combinatorial trades. We conclude the talk by introducing a new problem of designing popularity-change stable trades and provide constructions for the same based on specialized graph representations of popularity changes.
This is joint work with Charles Colbourn, Hoang Dau, Ryan Gabry, Chao Pan, and Jin Sima
Date and Time
Location
Hosts
Registration
- Date: 07 Nov 2024
- Time: 03:15 PM to 04:30 PM
- All times are (UTC+01:00) Stockholm
- Add Event to Calendar
- Lund University
- Klas Anshelms väg 10
- Lund, Skane lan
- Sweden
- Building: E-building
- Room Number: E:2517
- Contact Event Host
- Co-sponsored by Michael Lentmaier, Lund University
Speakers
Prof. Olgica Milenkovic of University of Illinois, Urbana-Champaign
Distributed storage systems with dynamic access/load balancing
The problem of designing codes for distributed storage systems has received significant attention due to its relevance to many practical applications and interesting analytical challenges.
Many relevant distributed coding schemes are based on combinatorial designs which ensure efficient content regeneration and system recovery. An example includes fractional repetition codes that are both conceptually simple and easy to implement.
One important problem mostly overlooked in coding for distributed storage pertains to access balancing (i.e., load balancing) where one requires that the number of access requests to servers both during repair and data download are close-to-uniform. To implement such a balancing scheme, one needs to take into account the popularities of the stored data chunks or files as these dictate the number of access requests received by the servers.
We introduce new access (load) balancing fractional repetition codes that represent the first instance of combinatorial designs with both classical intersection and labeling (coloring) constraints. The codes maintain the properties of classical fractional repetition codes but also ensure that the average popularity scores of data files at different servers are nearly the same. We then describe a dynamic popularity-change model and explain how it relates to combinatorial trades. We conclude the talk by introducing a new problem of designing popularity-change stable trades and provide constructions for the same based on specialized graph representations of popularity changes.
This is joint work with Charles Colbourn, Hoang Dau, Ryan Gabry, Chao Pan, and Jin Sima
Biography:
Olgica Milenkovic holds the Franklin W. Woeltge professorship in Electrical and Computer Engineering at the University of Illinois, Urbana-Champaign (UIUC). She obtained her MS degree in mathematics in 2001 and PhD in electrical engineering in 2002, both from the University of Michigan, Ann Arbor. Prof. Milenkovic's research interests lie in the areas of algorithm design and computing, bioinformatics, coding theory, machine learning and signal processing. Her scholarly contributions have been recognized by multiple awards, including the NSF Faculty Early Career Development (CAREER) Award, the DARPA Young Faculty Award, the Dean’s Excellence in Research Award, the University of Michigan ECE Distinguished Educator Award and several best paper awards. In 2013, she became a UIUC Center for Advanced Study Associate and Willett Scholar, while in 2015 she served as Distinguished Lecturer of the Information Theory Society. She became an IEEE Fellow in 2018. Prof Milenkovic has served as Associate Editor of the IEEE Transactions on Communications, the IEEE Transactions on Signal Processing, the IEEE Transactions on Information Theory, the IEEE Transactions on Molecular, Biological and Multi-Scale Communications and the Transactions on Machine Learning Research. In 2009 and 2020 she assumed the role of Guest Editor-in-Chief of the IEEE Transactions on Information Theory on Molecular Biology and Neuroscience and the Memorial Tribute to V. I. Levenshtein.
Address:United States