I think the meaningful values for caps will come from the distributed scheduler effort. That branch really needs to drive what is required from the services.
RE flags: do you mean "x=1,x=2,x=3,y=9,y=8,y=7" vs. "x:1,2,3;y:9,8,7;" ?
RE storing the capabilities: yes, storing it in the db is a possibility. Since many capabilities are not static (disk remaining, bandwidth usage, cpu, etc) I had concerns about the frequency of the updates and the added load on the db. But this could use the same periodic storage scheme as the fanout mechanism.
Todd, thanks for the feedback.
I think the meaningful values for caps will come from the distributed scheduler effort. That branch really needs to drive what is required from the services.
RE flags: do you mean "x=1,x= 2,x=3,y= 9,y=8,y= 7" vs. "x:1,2,3;y:9,8,7;" ?
RE storing the capabilities: yes, storing it in the db is a possibility. Since many capabilities are not static (disk remaining, bandwidth usage, cpu, etc) I had concerns about the frequency of the updates and the added load on the db. But this could use the same periodic storage scheme as the fanout mechanism.