[Suggestion] Launch process is slow (Initialize VDUs in parallel)

[Rizwan-Qamar]

1. Code is from AllocateresourceTask.java
2. VDUs should be started in parallel (Unless they depend on each other). Each VDU should have independent initialization cycle. Failure of one VDU should not effect (abort) the initialization of other VDUs.

    for (VirtualDeploymentUnit vdu : virtualNetworkFunctionRecord.getVdu()) {
      List<Future<List<String>>> ids = new ArrayList<>();
      VimInstance vimInstance = vims.get(vdu.getId());
      if (vimInstance == null) {
        throw new NullPointerException(
            "Our algorithms are too complex, even for us, this is what abnormal IQ means :" + "(");
      }
      vimInstance = vimRepository.findFirstById(vimInstance.getId());
      log.debug(
          "Allocating VDU: "
              + vdu.getName()
              + " to vim instance: "
              + vimInstance.getName()
              + " - id: "
              + vimInstance.getId());
      for (NFVImage image : vimInstance.getImages()) {
        log.trace("Available image name: " + image.getName());
      }
      ids.add(
          resourceManagement.allocate(
              vdu, virtualNetworkFunctionRecord, vimInstance, userData, keys));

      for (Future<List<String>> id : ids) {
        id.get();
      }
    }

[lorenzotomasini]

Hi,

I would like to change the content of this issue by splitting in two different.

1. allocate vdu in parallel
2. put vnfr to error if one / all vdu fail

so i will close this issue with regards to point 1 and i will open another one for what concerns point 2

if this ok?
@mpauls ?

[mpauls]

sounds fair

[lorenzotomasini]

opened new one, @Rizwan-Qamar please follow the new one (#101). I will close this.

[lorenzotomasini]

The fix was not working, need to reopen the enhancement