Please note that this event list is no longer being updated.
For current events, please visit our new site
Public talks of interest to the Data Systems Group are posted here, and are also
mailed to the
dsg-faculty,
dsg-grads,
dsg-friends
mailing lists.
Subscribe to one of these mailing lists to receive e-mail notification
of upcoming events.
Everyone is welcome to attend.
Information about older events can be found in the
event archive.
2018 Events
|
PhD Seminar:
|
Wednesday January 17, 12:30 pm, MC 5417
|
|
Speaker:
|
Brad Glasbergen
|
|
|
Title:
|
Title: Apollo: Learning Database Query Correlations for Predictive Caching
|
|
Absract:
|
The performance of modern database applications is increasingly
dependent on remote access latencies. The worldwide distribution of users
results in significant latency for clients that are distant from the
application's data center. Systems that reduce latencies by caching query
data closer to clients are becoming more popular but do not take advantage
of application-specific behaviour patterns. In this talk, I will present
Apollo, a system that learns database query patterns and exploits them to
enhance query performance via predictive caching. Experiments demonstrate
Apollo's efficacy as a predictive cache for geo-distributed database
applications.
|
|
PhD Seminar:
|
Wednesday March 14, 1:00 pm, MC 5417
|
|
Speaker:
|
Michael Abebe
|
|
|
Title:
|
Title: EC-Store: A Dynamic Distributed Erasure Coded Storage System
|
|
Absract:
|
Cloud storage systems typically choose between replicating or erasure encoding data to provide fault tolerance. Replication ensures that data can be accessed from a single site but incurs a much higher storage overhead, which is a costly downside for large-scale storage systems. Erasure coding has a lower storage requirement but relies on encoding/decoding and distributed data retrieval that can result in increased response times. In this talk I will present EC-Store, a dynamic distributed erasure coded storage system that significantly reduces data retrieval times when compared to replicated and erasure coded storage systems. EC-Store achieves these reductions in latency by making intelligent data access decisions and dynamically moving data in response to system load and access patterns.
|
|
PhD Seminar:
|
Wednesday July 18, 1:00 pm, DC 2585
|
|
Speaker:
|
Anil Pacaci
|
|
|
Title:
|
Experimental Analysis of Streaming Algorithms for Graph Partitioning
|
|
Absract:
|
Graph partitioning in streaming setting is the problem of splitting graph structured data over a cluster of machines in the context of distributed graph processing systems. Streaming algorithms for graph partitioning has recently gained attention due to its ability to scale very large graphs with limited resources. This study first characterizes streaming algorithms for graph partitioning based on the distribution model. We identify edge-cut and vertex-cut model and provide analysis of existing algorithms in literature. The main objective is to understand how the choice of graph partitioning algorithm affects the system performance, resource usage and scalability. To this end, we provide systematic empirical evaluation of existing algorithms on large, real-world graphs using two classes of workloads: offline graph analytics and online graph queries. In our study, we argue that "no one size fits all" and choice of graph partitioning algorithm depends on: (i) type and degree distribution of the graph, (ii) characteristics of the workload and (iii) requirements of the application.
|
This page is maintained
by
Ken Salem.
