1 / 14

PlanetLab Current Status

PlanetLab Current Status. Brent Chun Timothy Roscoe David Culler 8 th November 2002. Overview. PlanetLab is about 5 months old Development primarily here and Princeton Research topics at Berkeley only just emerging This talk is a overview of where we’re currently headed.

etana
Télécharger la présentation

PlanetLab Current Status

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. PlanetLab Current Status Brent Chun Timothy Roscoe David Culler 8th November 2002

  2. Overview • PlanetLab is about 5 months old • Development primarily here and Princeton • Research topics at Berkeley only just emerging • This talk is a overview of where we’re currently headed. • Slightly Mothy-centric since Brent is ill • But go talk to him!

  3. Deployment • 101 nodes, 42 sites • Line of University volunteer sites • Some Industrial research volunteers • Budget for colo centers • Boot CD in development for next phase

  4. Research Agenda • Very early stages • S/w development, operations and support moving (slowly) to DSL team • Current small-ish projects here: • Resource brokers • DHT benchmarking • InfoSpect monitoring • Spread-spectrum storage • Deployment languages • Economic resource management

  5. Resource Brokers / Slices • Brent, Mothy, Larry Peterson, etc. • IPTPS: “Slices: Providing a Stable Core for Peer-to-Peer Applications” • Key PlanetLab rsrc allocation issue • How does an application request a slice? • How is it allocated? • How does the process span multiple providers? • How are allocations verified?

  6. DHT Benchmarking • Mothy, Sean Rhea (Oceanstore) • IPTPS: “CANs Need Application-Driven Benchmarks” • Aims: • Design good benchmarks for DHTs/CANs/DOLRs • Get real results using PlanetLab • Demonstrate the differences between systems • Currently: Tapestry, Chord • Pastry, Strawman in process

  7. Rules InfoSpect Ping Ssh Results of Prolog queries Observations as Prolog facts Ganglia Discovery tools / wrapper scripts Prolog Knowledge Base Results

  8. InfoSpect monitoring • Mothy • SIGOPS EW: “InfoSpect: using a Logic Language for System Health Monitoring in Distributed Systems” • Use of Prolog engine for PlanetLab fault diagnosis • “Axioms” prime the system • Small “driver scripts” discover information • Inference rules encode assumptions • Queries return anomalous conditions

  9. InfoSpect example • Trivial example: badstuff(Problem,Site) :- db_host(Site,DbHost,DbIp,_,_), ssh_greeting(DbIp,_), ping_failed(DbIp), swritef(Problem, '%w (%w) is filtering pings (and possibly other stuff)', [ DbHost,DbIp ]). • About a dozen of these tests and growing. • Runs continuously • Somewhat depressing results 

  10. Spread-spectrum storage • Mothy + Steve Hand (Univ. Cambridge) • Remember interview talk  • Transient, “soft capacity” storage for PlanetLab • User software distribution • Log file / results collection • Other calls for storage services • Oceanstore, SFS,

  11. Deployment languages • Mothy, and whoever • Appropriate languages for expressing the deployment of PlanetLab applications • Apply declarative techniques • Borrow ideas from component languages • Possible synergy with macroprogramming • Just a wild and crazy idea right now

  12. Economic Resource Management • Brent, et.al. • Policies built over slice model • Resource consolidators • etc. • Lots of code being written • Please ask Brent when he’s recovered – if he were here he would say a lot more.

  13. Other work we know about: • UCB: Oceanstore, ROC • UWash scriptroute • Duke AC/DC (adaptive overlays) • Princeton: RON routing service • Cambridge: PlanetProbe • Rice: Topology-aware Pastry • CMU: launching DoS attacks • …and many more

  14. Summary • Still “finding our feet” research-wise • Lots of issues coming out of deployment • Beginning to engage UCB students more • Reasonably gaggle of research directions

More Related