Page tree
Skip to end of metadata
Go to start of metadata

  • Demo of MultiScheduler (Logan)
    • The MultiScheduler allows use to test that the schedulers (serial and parallel ) return the same ending result when provided the same set of batches. The MultiScheduler implements the same trait and the other schedulers, and also takes a collection of the sub-schedulers and a sub-scheduler handler. If all schedulers return the same result, the Multischeduler will return the result back as normal. If there is a mismatch, an error is returned. (Logan)
      • What is a sub-scheduler? (Dan)
        • It can be either a serial scheduler or a parallel scheduler. It can be any combination of schedulers (for example 100 serial schedulers) to verify that they all return the same result. 
      • Why would there be multiple schedulers? (Dan)
        • The point is to verify that the serial and parallel schedulers return the same result, but we can scale that up to verify them against each other. You would not use this for production (Shawn)
      • Does this have a visadd so it could be used as a scheduler? (Dan)
        • Yes, it looks like a regular scheduler, It has the same interface, so it could be used, but it has a lot of duplication so it would never be used for product. (Shawn)
      • Is there anything difficult about this scheduler concept for Fabric to adopt? (Dan)
        • For the scheduler, no, but the way Fabric uses state will require some sort of global lock  around state. (Shawn)
      • I assume the read set and set state would still allow use of the Parallel Scheduler? (Dan)
        • Yes, we would use their keys as natural keys. But we probably couldn't do any wildcarding. (Shawn)
  • Review of the Multi-Project Kanban Board for Transact (Mark)
    • What is the command family? (Arun)
      • It is a family that can be used for testing that sends commands to set state, delete state, get state, fail transaction etc.. This will be used for correctness and performance testing. (Peter)
    • Is it used for resuming receipts from another change? (Dan)
      • That is one way it can be used, to test that the same result of state is computed (state root hash). (Peter)
  • Open Forum
    • The Transact announcement blog was posted last Thursday.  If anyone has questions, please ask them in # transact in RocketChat(Mark)
    • Is there a lot that has to be done to the RFC repo before we can submit the new RFCs? (Peter)
      • Yes, there is more that needs to be done to set it up (governance readme, directory, etc) (Shawn)

  • No labels