LoPC (original) (raw)

LogGP: Incorporating Long Messages into the LogP Model - One step closer towards a realistic model for parallel computation

1995

We present a new model of parallel computation---the LogGP model---and use it to analyze a number of algorithms, most notably, the single node scatter (one-to-all personalized broadcast). The LogGP model is an extension of the LogP model for parallel computation [CKP + 93] which abstracts the communication of fixed-sized short messages through the use of four parameters: the communication latency (L), overhead (o), bandwidth (g), and the number of processors (P ). As evidenced by experimental data, the LogP model can accurately predict communication performance when only short messages are sent (as on the CM-5) [CKP + 93, CDMS94]. However, many existing parallel machines have special support for long messages and achieve a much higher bandwidth for long messages compared to short messages (e.g., IBM SP-2, Paragon, Meiko CS-2, Ncube/2). We extend the basic LogP model with a linear model for long messages. This combination, which we call the LogGP model of parallel computation, has o...

Validation and Application of the LoPC model

1997

The LogP model is useful for modeling and optimizing tightly synchronized parallel algorithms with regular, ordered communication patterns on active-message based systems. However, it does not account for the contention for message processing resources at a node. The LoPC model accounts for costs of contention for message processing. In this work, we validate the LoPC model’s predictions against measured values of runtime on the Wisconsin Cluster Of Workstation (COW). We study three message-passing parallel applications: Radix-sort, Em3d and Barnes-Hut. However, we could validate the model predictions only for Em3d. To model the specific characteristics of the COW, we extended the LoPC model to include different overheads for the request and response handlers. Also, to determine the applicability of the LoPC model for applications with different characteristics, we developed a set of microbenchmarks and compare their runtime on the COW with the LoPC model predictions.

International Journal of Parallel Programming manuscript No. (will be inserted by the editor) Fast LH∗

2016

Linear Hashing is an efficient and widely used version of extensible hash-ing. LH ∗ is its distributed version that stores key-value pairs on up to hundreds of thousands of sites in a distributed system. LH ∗ implements the dictionary data struc-ture efficiently by not using a central component and allows the key-based operations of insertion, deletion, actualization, and retrieval as well as the scan operation. Be-cause it does not use a central addressing component, clients or servers in LH ∗ can commit an addressing error by sending a request to a wrong server. This server then forwards the message to the correct server either directly or in one but never more than one additional forward operation. We discuss here methods to avoid this dou-ble forward, which, while rare, still might breach quality of service guarantees. We compare our methods with LH∗RSP2P that pushes information about changes in the file structure to clients, whether they are active or not. A second problem espe...