Search

Multiple sequence alignment (MSA), used in biocomputing to study similarities between different genomic sequences, is known to require important memory and computation resources. Nowadays, researchers are aligning thousands of these sequences, creating new challenges in order to solve the problem using the available resources efficiently. Determining the efficient amount of resources to allocate is important to avoid waste of them, thus reducing the economical costs required in running for example a specific cloud instance. The pairwise alignment is the initial key step of the MSA problem, which will compute all pair alignments needed. We present a method to determine the optimal amount of memory and computation resources to allocate by the pairwise alignment, and we will validate it through a set of experimental results for different possible inputs. These allow us to determine the best parameters to configure the applications in order to use effectively the available resources of a given system.

MMORPG (Massively Multiplayer Online Role Playing Games) is the most popular genre among network gamers, and now attract millions of users, who play simultaneously in an evolving virtual world. This huge number of concurrent players requires the availability of high performance computation servers. Additionally, gaming aware distribution mechanisms are needed to distribute game instances among servers to avoid load imbalances that affect performance negatively. In this work, we tackle the problem of game distribution and scalability by means of a hybrid Client-Server/P2P architecture that can scale dynamically according to the demand. To manage peak loads that occur during the game, we distribute game computation across the system according to the behavior of MMORPGs. We distinguish between the computation associated with the Main Game, that affects all players, and the computation of Auxiliary Games that affects only a few players and acts in isolation from the execution of the Main Game. Taking this distinction into account, we propose a mechanism that is focused in the distribution of Auxiliary Games, as an entity, across the pool of servers and peers of the underlying hybrid architecture. We evaluate the performance of the balancing mechanism taking the criteria of latency and reliability into account, and we compare the effectiveness of the mechanism with a classic approach that applies load balancing to individually players in a Client-Server system. We show that the balancing mechanism based on the latency criteria provides lower latency than the classical proposal, while in relation to reliability, we obtain a failure probability of under 0.9 % in the worst case, which is amply compensated by the scalability provided by the use of the P2P area. [ABSTRACT FROM AUTHOR]

The exploitation of parallelism in a message passing platform implies a previous modeling phase of the parallel application as a task graph, which properly reflects its temporal behavior. In this paper, we analyze the classical task graph models of the literature and their drawbacks when modeling message passing programs with an arbitrary task structure. We define a new task graph model called Temporal Task Interaction Graph (TTIG) that integrates the classical models used in the literature. The TTIG allows us to explicitly capture the ability of concurrency of adjacent tasks for applications where adjacent tasks can communicate at any point inside them. A mapping strategy is developed from this model, which minimizes the expected execution time by properly exploiting task parallelism. The effectiveness of this approach has been proved in different experimentation scopes for a wide range of message passing applications. [ABSTRACT FROM AUTHOR]