PlanetLab Overview
This article will be devoted to PlanetLab (http://www.planet-lab.eu) - a network widely used by scientists to test new network services or modify existing ones. Currently, the use of PlanetLab as a platform for testing is included in the list of standard research methodology for distributed systems. During its existence, PlanetLab has been used to test a large number of different services: p2p networks, including DHT 1 , anycast 2 , 3 , distribution of large file sizes 4 , network analysis 5 , diagnostics of anomalies and errors 6 , and many others.
The document describing the idea of ​​creating PlanetLab and its basic concepts was first published in 2002. Thanks to the sponsorship of Intel, these ideas were not in vain, and already in 2003 the first 100 nodes were raised in 40 different universities around the world. To date, the project is supported by many sponsors . Currently, PlanetLab consists of 1109 nodes distributed over 535 different points of the world.
')
In order to be able to use PlanetLab, you must be an employee of one of the organizations that host the PlanetLab nodes. There is an exception to this rule - free public services are deployed in the network: proxy , p2p Coral Content Distribution Network .
Proxy PlaneLab, by the way, should not be used due to anonymity problems - any activity on proxy is logged . The description of this fact in runet found here .
So, after registering and becoming a happy user of PlanetLab, a beginner will have to learn not only the API, with which you can manage resources, but also network terminology - what are slices, slivers, nodes, etc (see below). But once you get used to it, it becomes very easy to manage resources.
One distributed service should run on multiple machines. The user easily (or automatically, having programmed) selects the available network nodes that he wants to use. On these nodes, resources are immediately allocated to him, all information about the user is taken from the central database. So, once you download a public rsa key, you can immediately use it to connect to any newly added node via ssh. By the way, network traffic is not limited, it can be used quite intensively. This leads to the fact that the owners of the sites have to pay somewhere around $ 1000 per year for used traffic.
The Central Bootserver “ PlanetLab Central ” (PLC) allows centralized management of all PlanetLab resources, starting with the choice of operating system running on each node, choosing which user has access to which node, and ending with choosing which resources will be allocated to which services. Moreover, the settings are changed both via the web interface and through the specialized XML-RPC API. I gave an example of using such an API in this article . At the same time, site owners themselves do not have the ability to manage resources, including which services are running on their site.
This concept makes it easy for PlanetLab users to deploy their services for further testing. PlanetLab allows you to run various services on nodes in parallel, which is achieved using distributed virtualization. The basic abstractions used in PlanetLab are slice and sliver .
Slice - these are all resources that are assigned to an individual user. This includes a list of all the individual nodes that have been selected by the user for testing.
Thus, each individual user’s service is launched on a slice from shared resources. Slices are distributed over a group of nodes selected by the user. On each of these nodes, several services are running, and the number of such services is not limited.
To separate services from each other, sliver is used. These are virtualized resources of a separate server that are assigned to a specific user. It is implemented using Linux- VServers . Network resources are virtualized using VNET .
The picture below illustrates the principle of the architecture of PlanetLab, the central box symbolizes the PLC.
I hope you were interested. You can read more about PlanetLab on the web site itself.
1 Sean Rhea, Brighten Godfrey, Brad Karp, John Kubiatowicz, Sylvia Ratnasamy, Scott Shenker, Ion Stoica, and Harlan Yu. 2005. OpenDHT: a public DHT service and its uses. SIGCOMM Comput. Commun. Rev. 35, 4 (August 2005), 73-84. DOI = 10.1145 / 1090191.1080102 http://doi.acm.org/10.1145/1090191.1080102
2 FREEDMAN, MJ, LAKSHMINARAYANAN, K., AND MAZIERES, D. OASIS: Anycast for any service. In Proc. 3rd Symposium on Networked Systems Design and Implementation (NSDI 06) (San Jose, CA, May 2006).
3 BERNARD WONG, ALEKSANDRS SLIVKINS, EGS MERIDIAN: In In Proceedings of the ACM SIGCOMM Conference, Philadelphia, Pennsylvania (August 2005).
4 ANNAPUREDDY, S., FREEDMAN, MJ, AND MAZIERES, D. SHARK: Scaling file servers via cooperative caching. In Proceedings of the 2nd USENIX Symposium on Networked Systems Design and Implementation (NSDI '05) (Boston, MA, USA, May 2005).
5 SPRING, N., WETHERALL, D., AND ANDERSON, T. SCRIPTROUTE: A public internet measurement facility, 2002.
6 ZHANG, M., ZHANG, C., PAI, VS, PETERSON, LL, AND WANG, RY PLANETSEER: Internet pathway. In OSDI
PS If you do not have access to articles, but you want to read, use this proxy
The document describing the idea of ​​creating PlanetLab and its basic concepts was first published in 2002. Thanks to the sponsorship of Intel, these ideas were not in vain, and already in 2003 the first 100 nodes were raised in 40 different universities around the world. To date, the project is supported by many sponsors . Currently, PlanetLab consists of 1109 nodes distributed over 535 different points of the world.
')
Using PlanetLab
In order to be able to use PlanetLab, you must be an employee of one of the organizations that host the PlanetLab nodes. There is an exception to this rule - free public services are deployed in the network: proxy , p2p Coral Content Distribution Network .
Proxy PlaneLab, by the way, should not be used due to anonymity problems - any activity on proxy is logged . The description of this fact in runet found here .
So, after registering and becoming a happy user of PlanetLab, a beginner will have to learn not only the API, with which you can manage resources, but also network terminology - what are slices, slivers, nodes, etc (see below). But once you get used to it, it becomes very easy to manage resources.
One distributed service should run on multiple machines. The user easily (or automatically, having programmed) selects the available network nodes that he wants to use. On these nodes, resources are immediately allocated to him, all information about the user is taken from the central database. So, once you download a public rsa key, you can immediately use it to connect to any newly added node via ssh. By the way, network traffic is not limited, it can be used quite intensively. This leads to the fact that the owners of the sites have to pay somewhere around $ 1000 per year for used traffic.
Basic Principles and Architecture of PlanetLab
Centralized Resource Monitoring
The Central Bootserver “ PlanetLab Central ” (PLC) allows centralized management of all PlanetLab resources, starting with the choice of operating system running on each node, choosing which user has access to which node, and ending with choosing which resources will be allocated to which services. Moreover, the settings are changed both via the web interface and through the specialized XML-RPC API. I gave an example of using such an API in this article . At the same time, site owners themselves do not have the ability to manage resources, including which services are running on their site.
Decentralized Management
This concept makes it easy for PlanetLab users to deploy their services for further testing. PlanetLab allows you to run various services on nodes in parallel, which is achieved using distributed virtualization. The basic abstractions used in PlanetLab are slice and sliver .
Slice - these are all resources that are assigned to an individual user. This includes a list of all the individual nodes that have been selected by the user for testing.
Thus, each individual user’s service is launched on a slice from shared resources. Slices are distributed over a group of nodes selected by the user. On each of these nodes, several services are running, and the number of such services is not limited.
To separate services from each other, sliver is used. These are virtualized resources of a separate server that are assigned to a specific user. It is implemented using Linux- VServers . Network resources are virtualized using VNET .
The picture below illustrates the principle of the architecture of PlanetLab, the central box symbolizes the PLC.
I hope you were interested. You can read more about PlanetLab on the web site itself.
Links to articles:
1 Sean Rhea, Brighten Godfrey, Brad Karp, John Kubiatowicz, Sylvia Ratnasamy, Scott Shenker, Ion Stoica, and Harlan Yu. 2005. OpenDHT: a public DHT service and its uses. SIGCOMM Comput. Commun. Rev. 35, 4 (August 2005), 73-84. DOI = 10.1145 / 1090191.1080102 http://doi.acm.org/10.1145/1090191.1080102
2 FREEDMAN, MJ, LAKSHMINARAYANAN, K., AND MAZIERES, D. OASIS: Anycast for any service. In Proc. 3rd Symposium on Networked Systems Design and Implementation (NSDI 06) (San Jose, CA, May 2006).
3 BERNARD WONG, ALEKSANDRS SLIVKINS, EGS MERIDIAN: In In Proceedings of the ACM SIGCOMM Conference, Philadelphia, Pennsylvania (August 2005).
4 ANNAPUREDDY, S., FREEDMAN, MJ, AND MAZIERES, D. SHARK: Scaling file servers via cooperative caching. In Proceedings of the 2nd USENIX Symposium on Networked Systems Design and Implementation (NSDI '05) (Boston, MA, USA, May 2005).
5 SPRING, N., WETHERALL, D., AND ANDERSON, T. SCRIPTROUTE: A public internet measurement facility, 2002.
6 ZHANG, M., ZHANG, C., PAI, VS, PETERSON, LL, AND WANG, RY PLANETSEER: Internet pathway. In OSDI
PS If you do not have access to articles, but you want to read, use this proxy
Source: https://habr.com/ru/post/137177/
All Articles