GPU node should request multiple jobs at once

[rvencu]

I need the ability to request multiple jobs at once from the GPU node such as

client.newJob(16)

to get a bundle of 16 jobs. this can be individual or composed shards as per #10 i.e. either 16 individual jobs or less when composed jobs are among them, while 8 composed jobs would be the minimum response to the above request

I guess client.downloadShard() would become client.downloadShards() or client.downloadJob() as for GPU node it makes no sense the naming downloadShard

[TheoCoombes]

Similarly to my response on #10, I'll add this to my todo list - it doesn't seem too hard to achieve.

[rvencu]

for this client.invalidURL() we need a more elaborated protocol

• say I request 16 shards
• I try the download links, some fail
• I send invalidURL() for those and I need to get new shards instead
• looping the above until all 16 are successful or out of shards condition appears then continue to process them

I think this should be done at my end, if you provide a way to request a number of jobs (and limit that to 20 or so to avoid bad requests )

the problem to be solved here is, will you allocate 16 jobs at once to one client? when I send invalidURL I should specify shard id somehow

[rvencu]

something like client.invalidURL('1b9437eb5f5144b9aeeae13d85909bb2')