1 / 9

ECE 753: FAULT-TOLERANT COMPUTING

ECE 753: FAULT-TOLERANT COMPUTING. Kewal K.Saluja Department of Electrical and Computer Engineering HIGH Level Fault-Tolerance: Checkpointing and recovery (more slides) – Forward error recovery. Overview. Forward error recovery Summary. Global checkpoint state. Global checkpoint state.

erek
Télécharger la présentation

ECE 753: FAULT-TOLERANT COMPUTING

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. ECE 753: FAULT-TOLERANT COMPUTING Kewal K.Saluja Department of Electrical and Computer Engineering HIGH Level Fault-Tolerance: Checkpointing and recovery (more slides) – Forward error recovery

  2. Overview • Forward error recovery • Summary ECE 753 Fault Tolerant Computing

  3. Global checkpoint state Global checkpoint state Checkpointing: Dist. Sys. (contd.) • Checkpoint based recovery – Comm. induced • Place system wide constraints on message passing (communication pattern) and checkpointing to guarantee progression of recovery line • For example if within every checkpoint interval all messages received precede all messages sent, then the system is domino effect free. Such a message passing system will appear to be as follows ECE 753 Fault Tolerant Computing

  4. Checkpointing: Dist. Sys. (contd.) • Checkpoint based recovery – Comm. induced • Consistency of state can be guaranteed (no domino effect will take place) if we take a checkpoint before every non deterministic event (note this is a special case of what we saw before). The challenge here is to reduce the number of checkpoints • Generalization of the previous statement (all “receives” to precede “sends”) have been studied in literature ECE 753 Fault Tolerant Computing

  5. A P0 m1 m7 m0 m4 B P1 m2 m3 m5 m6 P2 C Checkpointing: Dist. Sys. (contd.) • Log based recovery - Pessimistic logging (contd.) • Example and logs • Logs of • P0 is {m0, m4, m7} • P1 is {m1, m3, m6} • P2 is {m2, m5} ECE 753 Fault Tolerant Computing

  6. Forward error recovery • Consider the following scheme for checkpoint based rollback recovery • Two processors P1 and P2 • At checkpoints P1 and P2 compare results and save state (checkpoints) • Error detected at such a comparison, causes the processors P1 and P2 to roll back • Note: if there were three processors, they could potentially mask the faulty process/processor ECE 753 Fault Tolerant Computing

  7. error P1 chk2 P2 compare spare Forward error recovery (contd.) • Roll forward – two processors and a spare chk1 ECE 753 Fault Tolerant Computing

  8. Forward error recovery (contd.) • Additional issues • If we had three processors to begin with why not use fault masking? • Think of more than one job and the number of processors required – a single spare can be spare for many pairs of jobs • What if a second fault occurs while processor P1 is continuing? • Use spare for one more period ECE 753 Fault Tolerant Computing

  9. Summary • Discussed checkpointing and logging issues at length ECE 753 Fault Tolerant Computing

More Related