Recovery%20Oriented%20Computing:%20Update

1
Recovery Oriented ComputingUpdate

• Armando Fox (in loco Patterson)Summer ROC
  Retreat, June 2002

2
Welcome and ROC Philosophy

• ROC philosophy (Peress Law)
• If a problem has no solution, it may not be a
  problem, but a fact not to be solved, but to be
  coped with over time Israeli foreign
  minister Shimon Peres
• Failures (hardware, software, operator-induced)
  are a fact recovery is how we cope with them
  over time
• Availability MTTF/MTBF MTTF / (MTTF MTTR) -
  rather than just making MTTF very large, make
  MTTR ltlt MTTF
• ROC Principles
• Isolation and partitionability gt redundancy
• Enable fault injection, output checking gt online
  monitoring verification
• Undo support
• Diagnostic support

3
Major ROC Areas

• Failure detection and diagnosis
• Pinpoint
• FIG
• Internet service failure causes
• Recovery techniques and Design-for-Recovery
• Recursive Restartability
• Making state-management tradeoffs explicit
  (QAPSL)
• Firm state from infirm components (RAINS)
• Designing for Undo theory and practice
• Benchmarking and measurement
• Dependability benchmarks for various applications
• End-user availability measurements on the Web
• Why Internet services fail
• Estimating the cost of downtime
• Availability in the PSTN

4
Recent Publications

• ROC Techniques and Tools
• A Utility-Centered Approach to Internet Services
  Design. George Candea, Armando Fox, in SIGOPS
  European Workshop
• FIG A prototype Tool for Online Verification of
  Recovery Mechanisms. P. Broadwell, N.Sastry,
  J.Traupman, D.Patterson, in SHAMAN workshop at
  ICS 2002
• Rewind, repair, replay 3 Rs to Dependability.
  A. Brown and D. Patterson, SIGOPS European
  Workshop
• Including the Human Factor in Dependability
  Benchmarks. A. Brown, L. Chung, D. Patterson.
  In DSN 2002 Workshop on Dependability
  Benchmarking.
• ROC Measurements
• Architecture, operation, and dependability of
  large-scale Internet services three case
  studies. D. Oppenheimer and D.A. Patterson.
  Submission to IEEE Internet Computing special
  issue on Global Deployment of Data Centers,
  February 2002. (Shorter version in SIGOPS
  European Workshop)
• Measuring End-User Availability on the Web
  Practical Experience. Matthew Merzbacher and Dan
  Patterson.
• Lessons from the PSTN for Dependable Computing.
  P.Enriquez, A.Brown, D.Patterson, in SHAMAN
  workshop at ICS 2002.
• Fault monitoring/diagnosis
• An Online Evolutionary Approach to Internet
  Services. E. Kiciman, M. Chen, E. Brewer. In
  SIGOPS European Workshop

5
Recent Evangelism

• Evangelism publications
• Case for ROC Technical Report
• Introduction to Dependability (login)
• A Simple Way to Measure Cost of Downtime (LISA
  02)
• Evangelism talks
• Microsoft Research
• HPCA 02 keynote (Patterson)
• FAST keynote (Filesystems And Storage
  Technologies)
• IBM Autonomic Computing workshops (Almaden TJ
  Watson)

6
About ROC Retreats

• Purpose of semi-annual retreats
• Progress reports/talks from academia and industry
• Exposure/feedback on new ideas or work in
  progress
• Brainstorming in immersive atmosphere
• Industry/visitor feedback, opportunities for
  collaboration
• Water fights during rafting trip
• Logistics
• Web server with retreat talks/papers - thanks to
  Mike Howard and Bob Miller - http//172.16.10.43/r
  oc, WaveLAN ANY

7
Retreat Schedule - a work in progress

• Rest of today
• OceanStore update from Kubi
• Intros
• ROC talks
• All day Posters (especially right before after
  dinner)
• Tomorrow
• Morning OceanStore talks
• Afternoon Lunch and rafting
• Post-rafting breakout sessions followed by
  dinner
• Breakout reporting/joint panel session with
  SAHARA
• Wednesday
• Industry talk(s)
• Open mike/outrageous ideas session?
• Visitor feedback

8
Breakout Sessions

• Target 3-4 breakouts
• Using virtual machine technology for ROC
• Ideas for the second ROC showcase application
• Applying ROC to OceanStore
• Management and Self-healing of large-scale
  systems
• Is gt100 year storage a pipe dream?
• Other topics solicited
• Final breakout topics will be decided based on
  interest in each topic and limiting each group
  size