CLOUD COMPUTING: ESTIMATION OF EXECUTION DURATION PERTINENT TO CONCURRENT EXECUTION OF THREADS

摘要

Platform as a service (PAAS) in cloud computing is utilized to host app forexecutinginstructions/functions on sequences of bytes fetched from a cloud repository.The deployed appcan be scaled by increasing/decreasing the number of instances. Cloudproviders utilizeconsumed CPU time to increase/decrease the number of instances. In employedmethod, afraction of available CPU time is assigned to app. Consumption of CPU timeabove the thresholdincreases the number of instances and consumption of CPU time below thethreshold decreasesthe number of instances.A method to calculate known values of a quadratic equation for estimatingexecution time ofexecuting instructions/functions on sequences of bytes is devised. Thequadratic equation hasapplication in estimating required computing time, horizontal/vertical scalingof instances hostingan app, measuring consumed computing time for app's users and defining aquality of servicepertinent to execution duration.Sequential executing of instructions/functions starts by fetching a sequenceof bytes from a cloudrepository, proceeds by executing instructions/functions on the sequence ofbytes, and ends bystoring the result (sequence of bytes) in a cloud repository. By dividing thecode block ofsequential instructions/functions into sections, each section is executed in athread. A threadfetches sequence of bytes from a cloud repository, executesinstructions/functions on thesequence of bytes, stores sequence of bytes in a cloud repository, ortransports a sequence ofbytes to alternative instance.Threads performing similar task are executed concurrently. Execution durationof the threads isestimated utilizing a quadratic equation: a ×(Metric1)2 + b ×Metric' + c + d x (Metric2)2 +e × Metric2 + .function. . Known values of the quadratic equation arecalculated utilizing regressionanalysis to relate a number of metrics to execution durationmeasured/calculated for single-threading and multi-threading.