Application-Arrival Rate Aware Distributed Run-Time Resource Management for Many-Core Computing Platforms

摘要: Modern many-core computing platforms execute a diverse set of dynamic workloads in the presence varying application arrival rates. This inflicts strict requirements on run-time management to efficiently allocate system resources. On way towards kilo-core processor architectures, centralized resource approaches will most probably form severe performance bottleneck, thus focus has been turned study Distributed Run-Time Resource Management (DRTRM) schemes. In this article, we examine behavior DRTRM applications with malleable characteristics against stressing incoming interval rate scenarios, using Intel SCC as target system. We show that allocation is highly affected by input and propose an application-arrival aware framework implementing effective admission control strategy carefully utilizing voltage frequency scaling parts its infrastructure. Through extensive experimental evaluation, quantitatively analyze introduced scheme it achieves up 44 percent gains while consuming 31 less energy, comparison state-of-art solution. RTRM, respective metric values rise 62 45 energy gains, respectively.