Using Discrete-Event Simulation for Evaluating Distributed Systems
System modeling and simulation allow the designers of complex distributed systems to evaluate and predict the response time, throughput, and system availability so they can make informed design decisions before the actual implementation. In addition, simulation facilitates system evaluations for many different workloads, computational and network capacities and various failure scenarios at very large scale, without the high cost of infrastructure.
The main challenge in developing system simulators is determining the granularity and fidelity level of the simulation. Furthermore, choosing the right type of simulation and development tools and determining the inputs to the simulator are of great importance. For example, the inputs can come from probability distributions estimated based on measured real-world data or from the output of another independent simulation.
In this talk, I will discuss strategies for designing and developing a simulator for a multi-tier distributed system. I will illustrate this by sharing our experience at Oracle in developing a distributed storage system simulator using SimPy, an open-source Python library for discrete-event simulation.
Date and Time
Location
Hosts
Registration
- Date: 25 Sep 2018
- Time: 04:00 PM to 05:00 PM
- All times are (UTC-05:00) Central Time (US & Canada)
-
Add Event to Calendar
- Rice University, 6100 Main
- Electrical and Computer Engineering
- HOUSTON, Texas
- United States 77005
- Building: Duncan Hall
- Room Number: 1064
- Contact Event Host
-
https://events.rice.edu/#!view/event/event_id/13225
- Co-sponsored by Rice University, ECE Dept. - Prof. J. Cavallaro
Speakers
Aida Vosoughi of Oracle, Inc.
Using Discrete-Event Simulation for Evaluating Distributed Systems
System modeling and simulation allow the designers of complex distributed systems to evaluate and predict the response time, throughput, and system availability so they can make informed design decisions before the actual implementation. In addition, simulation facilitates system evaluations for many different workloads, computational and network capacities and various failure scenarios at very large scale, without the high cost of infrastructure.
The main challenge in developing system simulators is determining the granularity and fidelity level of the simulation. Furthermore, choosing the right type of simulation and development tools and determining the inputs to the simulator are of great importance. For example, the inputs can come from probability distributions estimated based on measured real-world data or from the output of another independent simulation.
In this talk, I will discuss strategies for designing and developing a simulator for a multi-tier distributed system. I will illustrate this by sharing our experience at Oracle in developing a distributed storage system simulator using SimPy, an open-source Python library for discrete-event simulation.
Biography:
Aida Vosoughi is a senior member of technical staff at Oracle. She is a member of the advanced development team where she works with the product teams to prepare new technologies for incorporation into Oracle’s products. Her current focus is designing and developing high level simulation tools for distributed database systems to evaluate and optimize for high performance, availability and efficiency. Aida received her Ph.D. in Electrical Engineering from Rice University in 2016 where she did research on distributed trust management in cognitive radio ad hoc networks and also hardware design for high throughput error correction codes in wireless communication systems. Aida received her M.S. in Electrical Engineering from North Dakota State University in 2010. She completed her B.S. and M.S. degrees in computer engineering at Amirkabir University of Technology in 2006, and 2008, respectively. Aida’s home page: vosoughi.info
Agenda
Talk at 4pm in Duncan Hall room 1064