Context Navigation

Changes between Version 2 and Version 3 of LowLatency

-                      v2
+                      v3
 Now suppose that each host requests work every Z seconds. Assume Z is small enough so that at least M requests arrive in any given period of length W. (TODO: figure out what this is, given a Poisson arrival process).
+Then, within W second of the batch creation, all of the jobs have been sent to hosts, and within T seconds (assuming no errors or client failures) they have been completed and reported. Note: this is a simplistic analysis, and doesn't take into account multiprocessors, hosts of different CPU speed, the possibility of sending multiple jobs to one client, the ability for Z to vary between hosts, and probably many other factors. If someone wants to analyze this in more generality, please do!
+Then, within W seconds of the batch creation, all of the jobs have been sent to hosts, and within T seconds (assuming no errors or client failures) they have been completed and reported. Note: this is a simplistic analysis, and doesn't take into account multiprocessors, hosts of different CPU speed, the possibility of sending multiple jobs to one client, the ability for Z to vary between hosts, and probably many other factors. If someone wants to analyze this in more generality, please do!
 == How to do low-latency computing ==
 …
 If you're interested in helping add these features to BOINC, please [ProjectPeople contact us].