VMware Tech Talk: Virtualization Anyware

Speaker: Tariq Magdon-Ismail , VMWare

Date: Thursday, October 17, 2013

Time: 6:00 PM to 8:00 PM Note: all times are in the Eastern Time Zone

Refreshments: 6:00 PM

Public: Yes

Location: G449

Event Type:

Room Description:

Host: Sonya Kovacic, CSAIL

Contact: Sonya Kovacic, skovacic@csail.mit.edu

Relevant URL: http://www.vmware.com/

Speaker URL: None

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu

Reminder Subject: TALK: VMware Tech Talk: Virtualization Anyware

"Enabling Hadoop for the Cloud: virtualization, data-management and storage optimization"

A prerequisite to architecting any cloud service is to have an elastic platform that can dynamically respond to the varying demands placed on the service by multiple tenants while at the same time meeting its SLAs. Ideally, such a platform would also optimize for the most efficient use of the underlying infrastructure. What this means to a framework like Hadoop is that in order to offer it up as a multi-tenant, fully isolated service, we need to be able to expand/shrink its cluster size (capacity and footprint) dynamically. However, because the Hadoop compute and data layers are tightly coupled, expanding or shrinking in an online manner, in response to some sudden surge in demand, is a challenge. In this talk we examine the nature of Hadoop and the hurdles that need to be overcome in order to turn it into a well performing elastic service. We explore some solutions using existing technologies and provide experimental data that shows we can achieve our goal with acceptable levels of performance. We further characterize the remaining overheads and present novel solutions to significantly optimize performance even further.

Speaker Bio:
Tariq Magdon-Ismail is a Senior Engineer at VMware. His current research interests sit at the intersection of systems virtualization and Big Data performance.

Prior to joining VMware, Tariq spent over 10 years at Sun Microsystems working on large-scale systems performance and scalability. Tariq served on the Sun SPARC hypervisor architecture committee where he led the development of the memory management specification. Recognized as a performance expert, he drove the development of several technologies for the Solaris operating system and UltraSPARC processor architecture.

Food will be provided!

Research Areas:

Impact Areas:

This event is not part of a series.

Created by Sonya Kovacic at Monday, September 30, 2013 at 12:00 PM.