The focus of this paper is on analyzing the effectiveness of SMP (Symmetric Multi-Processor) architecture for implementing Three-Tier Web-Servers. In particular, we considered a workload based on the TPC-W benchmark to evaluate the system. As the major bottleneck of this system is accessing memory through the shared bus, we analyzed what are the benefits of adopting several solutions aimed at boosting the global performance of the Web Server. Our aim is also to quantify the scalability of such a system and suggest solutions to achieve the desired processing power. The analysis starts from a reference case, and explores different architectural choices as for cache, scheduling algorithm, and coherence protocol in order to increase the number of processors possibly connected through the shared bus. Our results show that such an SMP based server could be scaled (up to 20 processor) quite above the limits expected for this kind of architecture, if particular attention is used in solving problems related to the process migration.
Foglia, P., Giorgi, R., Prete, C.A. (2002). Boosting the Performance of Three-Tier Web Servers Deploying SMP Architecture. In LNCS. LECTURE NOTES IN COMPUTER SCIENCE (pp.134-146).
Boosting the Performance of Three-Tier Web Servers Deploying SMP Architecture
GIORGI, ROBERTO;
2002-01-01
Abstract
The focus of this paper is on analyzing the effectiveness of SMP (Symmetric Multi-Processor) architecture for implementing Three-Tier Web-Servers. In particular, we considered a workload based on the TPC-W benchmark to evaluate the system. As the major bottleneck of this system is accessing memory through the shared bus, we analyzed what are the benefits of adopting several solutions aimed at boosting the global performance of the Web Server. Our aim is also to quantify the scalability of such a system and suggest solutions to achieve the desired processing power. The analysis starts from a reference case, and explores different architectural choices as for cache, scheduling algorithm, and coherence protocol in order to increase the number of processors possibly connected through the shared bus. Our results show that such an SMP based server could be scaled (up to 20 processor) quite above the limits expected for this kind of architecture, if particular attention is used in solving problems related to the process migration.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/11365/46876
Attenzione
Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo