Abstract | ||
---|---|---|
Recent developments in Big Data are increasingly focusing on supporting computations in higher data velocity environments, including processing of continuous data streams in support of the discovery of valuable insights in real-time. In this work we investigate performance of streaming engines, specifically we address a problem of identifying optimal parameters that may affect the throughput (messages processed/second) and the latency (time to process a message). These parameters are also function of the parallelism property, i.e. a number of additional parallel tasks (threads) available to support parallel computation. In experimental evaluation we identify optimal cluster performance by balancing the degree of parallelism with number of nodes, which yield maximum throughput with minimum latency. |
Year | Venue | Field |
---|---|---|
2016 | ADBIS (Short Papers and Workshops) | Data stream mining,Latency (engineering),Computer science,Degree of parallelism,Parallel computing,Thread (computing),Throughput,Big data,Computation |
DocType | Citations | PageRank |
Conference | 1 | 0.41 |
References | Authors | |
7 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Nigel Franciscus | 1 | 1 | 0.41 |
Zoran Milosevic | 2 | 548 | 54.38 |
Bela Stantic | 3 | 198 | 38.54 |