Lightning Talks (Scalable Software)
October 4, 2013
10:15 am–11:15 am
Location: MCC: 101 H-J
Finding Performance Problems Automatically with Feedback-Directed Learning Software Testing (10 minutes)
Presenter: Qi Luo, College of William and Mary
Absract: A fundamental question of performance testing is how to select a manageable subset of the input data faster for finding performance problems in applications automatically. We propose a solution, that extracts rules from execution traces of applications, and then uses these rules to select test input data for finding performance problems automatically. This adaptive feedback-directed learning testing system is implemented, and we applied it on two open-source applications.
Exploring the design space of processing-in-memory architecture for exascale computing (10 minutes)
Presenter: Dong Ping Zhang, Advanced Micro Devices
Abstract: This talk highlights the growing importance of co-design of software solutions and hardware architectures in general purpose computing industry. It also explores why computational capacity as a single metric is out-of-date and finally why power consumption of communication and data movement are as important.To provide an example, this talk focuses on exploration of processing-in-memory (PIM) design space through the software design solution of a PIM API, simulator and performance models.
One to Many: A Database Scaling Story (10 minutes)
Presenter: Tamar Bercovici, Box
Abstract: The database is notorious for being one of the most challenging web-components to scale. Come learn about how we transitioned Box from a single relational MySQL database, to a scalable distributed architecture called sharding, which spans multiple MySQL servers. We’ll discuss why we chose sharding as our scaling solution, our design process, and our incremental approach for transitioning to this architecture, all the while serving billions of queries per day.
Set for Success: Product Development Practices that Enabled Facebook to Reach One Billion Users (10 minutes)
Presenter: Julia Lee, Facebook
Abstract: How did Facebook grow to one billion users and what could be learned? This presentation describes scalable product development and software engineering practices that allowed Facebook to reach one billion users.
We will discuss best practices around understanding our users and identifying highest potential features. Discipline around instrumentation, data analysis, process, and decision-making will also be highlighted.
Centiman: A Scalable Online Transaction Processing System (10 minutes)
Presenter: Bailu Ding, Cornell University
Abstract: Online transaction processing (OLTP) systems lie at the backbone of our digital economy. However, OLTP systems are notoriously hard to scale with the transaction load. We propose Centiman, a novel high performance scalable distributed OLTP system based on optimistic concurrency control with novel parallel validation and memory management. On the industry-standard TPC-C benchmark, Centiman can scale to up to 250,000 transactions per second on a commodity cluster.
Design Principles for High Scale Distributed Systems (10 minutes)
Presenter: Irina Frumkin, Microsoft
Abstract: Building high scale, fault tolerant distributed systems to handle the needs of today’s businesses is no longer an option; it’s a must have. In this talk I plan to cover core building blocks for creating scalable distributed systems. I will share my teams learning’s, pitfalls and experiences with trying to get our distributed system to scale to 4.3 million messages per second.