Resilient Systems — Part One: Enterprise Grade Storage for the Price of Bare Metal
Brown Bag Lecture by Michael Kushnir | 11/1/2011 11AM-12PM | 7th Floor Conference Room, Bldg 38A
Abstract: In the age of open source everything and 99.99% uptime guarantees, highly-available fault-tolerant system architectures have become a standard rather than a luxury. The ongoing global recession has led to declining revenues, and down-time due to IT maintenance or failure is no longer an acceptable expense for many organizations. At the same time, declining IT budgets are quickly moving expensive proprietary fault-tolerance technologies out of the reach of organizations with small-business budgets. However, systems with multiple layers of fault tolerance are no longer limited to specialty systems. Thanks to free operating systems and open source software, enterprise-grade fault tolerant systems can now be built for the cost of bare metal and deployed within hours. This two part lecture series will discuss tools modern tools and implementation scenarios for building resilient systems that can sustain multiple component, server, and service failures without incurring downtime focusing specifically on shared-nothing cluster architectures based on free open source software. The lecture will discuss recent experiences in the design and implementation of the OPENi virtualized distributed computing architecture and other examples.
This first lecture will focus on storage– the backbone of any reliable information system. If storage fails, the system fails. This lecture will present several architectures with enterprise-grade reliability that can built on commodity-grade hardware and free open source software. The talk will cover equipment options, RAID configuration options, and storage server deployment scenarios. Windows based architectures will be covered briefly, but the main focus will be Linux based free and open source solutions.
Bio: Michael Kushnir joined the Communications Engineering Branch of the Lister Hill National Center for Biomedical Communications as a system architect and administrator in November, 2010. Michael specializes in RHEL/CentOS and VMware vSphere administration. Prior to NLM, Michael was head administrator for internet startups AdGlobalMedia.Com and YouSeeAds.Com where his work focused on low-cost, fully-redundant, and infinitely-scalable systems. Michael has fifteen years of experience in Windows administration and five years in Linux administration working mostly with free RHEL clones such as CentOS, OpenFiler, and etc…
Michael received a B.S. degree in Management Science from Johns Hopkins University in 2004 and a J.D. from the University of Baltimore School of Law in 2010. During his academic pursuits, Michael has conducted primary research on the feasibility of slots in Maryland for the Maryland House of Delegates Ways and Means Committee, and on banking law and regulation in the U.A.E. University of Baltimore Center for International and Comparative Law.