This preview shows pages 1–3. Sign up to view the full content.
This preview has intentionally blurred sections. Sign up to view the full version.View Full Document
Unformatted text preview: Distributed System Management: PlanetLab Incidents and Management Tools Robert Adams Intel Corporation PDN03015 November 2003 Status: Ongoing Draft. Distributed System Management: PlanetLab Incidents and Management Tools Robert Adams Intel Corporation 4 November 2003 Abstract PlanetLab is an open, global, distributed test bed for developing, deploying and accessing planetary-scale network services. Its goal is to be the infrastructure for a new generation of applications and services in the Internet. Supporting this new generation of distributed applications and services in the open Internet presents new challenges in support, maintenance and administration. This paper describes some of the management incidents that occurred in the first year of PlanetLab operation and describes some of the technologies that have been developed to cope with these problems with an eye to exploring the requirements for scalable management of an open distributed computer system. 1 PlanetLab Operation PlanetLab is an open, global, distributed test bed for developing, deploying and accessing planetary- scale network services [PLANETLAB]. As of October 2003, there are more than 200 machines at 90 sites worldwide 1 available to support both short-term experiments and long-running network services. Over the next few years, PlanetLab will grow to over 1000 nodes and host some of the newest and innovative services available. A small support team monitors the operation of all the PlanetLab nodes and fields support emails from users. The support team fulfills several tasks: keeping PlanetLab running; supporting the installation and operation of new PlanetLab nodes; fielding support questions and problems; developing basic PlanetLab user features; and developing tools for management and tracking of PlanetLab operation. To better understand how PlanetLab is run, what follows is a brief description of how PlanetLab operates. PlanetLab is a collection of computers distributed around the Internet. The individual computers are called nodes . PlanetLab nodes run a PlanetLab PDN: PDN-03-015. http://www.planet- lab.org/pdn/pdn-03.015.pdf . 1 Status and size of PlanetLab is available at http://www.planet-lab.org/. standard version of Linux with some enhancements for supporting multiple applications and users. Applications share a PlanetLab node by residing in separate virtual servers. To each application, the running environment looks like a private Linux computer a user has root access to the system files and can install Linux packages (RPMs, etc) as required. The system kernel has been modified to use vservers [VSERVER] for each user of the node. Vservers are akin to the BSD jail code and gives each user of a node its own separate copy of the system files and certain special privileges that make it look like the user has complete control of the computer....
View Full Document
- Spring '08