Scheduling Metric-Space Queries Processing on Multi-Core Processors

Abstract

Abstract-This paper proposes a strategy to organize metricspace query processing in multi-core search nodes as understood in the context of search engines running on clusters of computers. The strategy is applied in each search node to process all active queries visiting the node as part of their solution which, in general, for each query is computed from the contribution of each search node. When query traffic is high enough, the proposed strategy assigns one thread to each query and lets them work in a fully asynchronous manner. When query traffic is moderate or low, some threads start to idle so they are put to work on queries being processed by other threads. The strategy solves the associated synchronization problem among threads by switching query processing into a bulk-synchronous mode of operation. This simplifies the dynamic re-organization of threads and overheads are very small with the advantage that the overall work-load is evenly distributed across all threads.

Accessing very high dimensional spaces in parallel

“…We compared our two parallel search algorithms with a parallel version of the search method that uses the LC index implemented using OpenMP [9] on shared memory multiprocessor sun. To assess the performance of the algorithms we took as reference an OpenMP parallel implementation of the brute force method.…”

“…There are several parallel algorithms that use the LC index on different kinds of parallel platforms: distributed memory [11] and shared memory [9] multiprocessors, GPUs [6] and multi-GPU platforms [7]. We will compare our parallel algorithms with the local version of the algorithm introduced in [9] to deal with high query traffic.…”

“…We will compare our parallel algorithms with the local version of the algorithm introduced in [9] to deal with high query traffic. This algorithm is implemented with OpenMP and uses inter-query parallelism to distribute the processing of different queries on different cores of a shared memory multiprocessor.…”

Range query processing on single and multi GPU environments

“…As in some previous papers [11][12][13] we assume a high frequency of incoming queries and exploit coarse-grained inter-query parallelism, i.e. we always solve nq queries in parallel.…”

“…As a basis we have taken the proposals for implementation on multi-core systems introduced on [13]. In the so-called Local method, queries are distributed across threads following a round-robin distribution.…”