Improving scalability of Bag-of-Tasks applications running on master–slave platforms

Scalability analysis of embarassingly parallel applications on large clusters

“…More recently [3] we proved that isoefficiency and isospeed are equivalent for BoT applications executing on homogeneous master-slave platforms, considering the same application model and execution model used in this work. Therefore we consider both metrics equivalent for the models presented in this extended abstract.…”

“…As pointed out by [6], the one-port model is nicely suited to LAN network connections. Other communication models are considered in [3] [4].…”

“…Despite the fact that lowerbound functions represent best case scenarios, they are achievable for some instances of embarrassingly parallel applications. For details see [3] [4].…”

“…Several theorems are stated but, due to space limits, all proofs are omitted. For detailed proofs please consult [3] [4]. This paper is organized as follows: the models considered in this work are presented in section II.…”

A Grid Broker Pricing Mechanism for Temporal and Budget Guarantees

“…This ensures resource providers are not going to either undervalue or overvalue resources, which in turn can lead to unrealised profit or utility [17,21]. The models in [23,14,5,1] all make little attempt to cope with data transfer requirements, which has been shown by [27,18] to be a substantial problem for the scalability of algorithms. We intend to add pricing based on data transfer to our model and to account for transfer times and costs.…”