Time+Place: Thursday 15/12/2016 14:30 Room 337 Taub Bld.
Title: Cloud resource provisioning for social notifications at mass scale
Speaker: Roman Vitenberg - COLLOQUIUM LECTURE http://folk.uio.no/romanvi/
Affiliation: Dept. of Informatics, University of Oslo
Host: Roy Friedman

Abstract:

Delivery of social notifications has become a popular component in many 
electronic services. While a small number of services for social 
notifications, e.g., Twitter, are dedicated, an increasingly large
number of such services are provided as an enhancement of the main 
functionality. For example, the main music streaming service of Spotify 
is accompanied by an engine that delivers real-time notifications to 
each online user. These notifications report when a user's friend comes 
online and goes offline, what music track a user's friend listens to, 
when a user's friend updates a shared content, such as a playlist, etc. 
Similar services for social notifications are provided alongside most 
real-time cooperative activities: video and audio streaming, mobile and 
ad-hoc interaction, online gaming, large-scale human gatherings during 
the commute, Olympic games, exhibitions, as well as many others. The 
volume of social notifications is quite staggering. For example, the 
notification service at Spotify delivers over a billion notifications 
every single day, with a total volume of over 2 terabytes. In this talk, 
I will provide a brief analysis of the Spotify data, with the aim to 
facilitate design and validation of new solutions.

Traditionally notification services have been deployed on in-house 
enterprise datacenters. However, with the advent of cloud computing, a 
viable alternative of running such services in the cloud became 
available. In this talk, I will present a few key challenges of delivery 
semantics as well as resource provisioning for deployment of 
notification services, define corresponding optimization problems, and 
describe our solutions to these challenges. We evaluate the proposed 
solutions experimentally using real traces from Spotify and Twitter 
along with the Amazon EC2 pricing model. The solution runs in under 30 
seconds for the Spotify workload with 5 million subscribers and 1.1 
million topics and under 25 minutes for the Twitter workload with 30 
million subscribers and 8 million topics, using a commodity hardware.

The content of the talk is based on publications in the DEBS 2013, 
Infocom 2014, and ICDCS 2014 conferences.

Short Bio:
==========

Roman Vitenberg is a Professor at the Department of Informatics at the 
University of Oslo. He received his PhD in Computer Science from the 
Technion - Israel Institute of Technology. In the past, he was a 
visiting researcher at the University of California at Santa-Barbara, 
Universita di Roma La Sapienza, and Universidad Politecnica de Madrid. 
He also spent three years as a research staff member at IBM Research 
where he contributed to the design of the high-availability component of 
WebSphere.
His research interests are broadly in the area of distributed 
applications, middleware and algorithms; including specification, 
design, analysis, implementation, performance evaluation, and software 
engineering. In particular, he has been working on large-scale 
communication, object-oriented and component-based platforms, 
distributed event-based systems, consistency models, and fault-tolerant 
distributed computing. He is recipient of best paper awards at 
ACM/IFIP/USENIX Middleware, ACM SAC, and ACM DEBS conferences.

===========================================
Refreshments will be served from 14:15
Lecture starts at 14:30