Bounds on the Client-Server Incremental Computing

Bounds on the multi-clients incremental computing for homogeneous decreasing computation sequences

“…Coordinating computations with communications to achieve high performance computing for a single task was investigated in [5]. The bounds on the time complexity for employing incremental computing to hide communication latency have been studied in [6]. Our research extends the work in [5].…”

“…When the tasks can be performed much more quickly using special hardware or software environment, the tasks are assigned to the server that can easily finish the tasks. In [6], incremental computing has been defined as the technique that the server employs any available CPU cycles to compute partial results based on the current partial input. The number of triggered computations depends on the data items packed in a message and its order of arriving at the server.…”

On the assessment of input streams for incremental network computing

“…In the paper, the strategies of scheduling messages for maximizing the utilization of a networked computing device were studied. In [5], the authors studied whether it is possible to find a best input stream for an existing algorithm, which maximizes the processor utilization of the computing device for any traffic pattern at any time. In the paper, best input streams have been found for three applications: the product of two polynomials, matrix multiplication and Fast Fourier Transform (FFT).…”

“…It can be achieved by partitioning the input data into several data pieces, then pipelining the data pieces into the network. In incremental computing mode [5], the computing device starts its computations using the partially arrived data to aggressively produce the partial result. However, network latency cannot be effectively hidden without considering the order of sending data [4].…”

“…the total length of a path. It is known that the problem of input stream finding is NP-Hard in the strong sense [5]. Heuristic search algorithms are often employed to solve the class of problems.…”