Fine-grain temperature monitoring for many-core systems

Abstract

The power density may limit the amount of energy a many-core system can consume. A many-core at its maximum performance may lead to safe temperature violations and, consequently, result in reliability issues. Dynamic Thermal Management (DTM) techniques have been proposed to guarantee that many-core systems run at good performance without compromising reliability. DTM techniques rely on accurate temperature information and estimation, which is a computationally complex problem. However, related works usually abstract the temperature monitoring complexity, assuming available temperature sensors. An issue related to temperature sensors is their granularity, frequently measuring the temperature of a large system area instead of a processing element (PE) area. Therefore, the first goal of this work is to propose a finegrain (PE level) temperature monitoring for many-core systems. The second one is to present a dedicated hardware accelerator to estimate the system temperature. Results show that software performance can be a limiting factor when applying an accurate model to provide temperature estimation for system management. On the other side, the hardware accelerator connected to the many-core enables the fine-grain temperature estimation at runtime without sacrificing system performance.

Machine Learning based Electromigration-aware Scheduler for Multi-core Processors

“…The lifetime of the cores when operating with the frequencies estimated by the algorithm is having a maximum deviation of 2.85 % from the required lifetime. Alzemiro et al [21] 0.020 o C Kaicheng Zhang et al [32] 2.900 o C Carlton Knox et al [34] 1.390 o C Using the methodology proposed, embedded application developers can perform a fast design space exploration between the lifetime reliability of the processor cores and the performance requirement of the tasks. Fig.…”

“…Multiple scheduling schemes based on the efficient and simple thermal model are used [20] for managing the operations of homogeneous processor platforms. Thermal Estimation Accelerator (TEA) [21], a processing element level monitoring scheme for the temperature at runtime using hardware accelerators can serve as a benchmark for Dynamic Thermal Management (DTM) methods.…”

Mapping and Migration Strategies for Thermal Management in Many-Core Systems

“…Due to the complexity of thermal models, system characterization and application profiling compose the temperature estimation for replacing thermal monitoring in thermal management [8,9]. Therefore, thermal management strategies usually do not include reactive actuation because they would impact on temperature estimation [14]. For example, task migration is a management strategy to deal with traffic issues and perform occasional task remapping with low overhead [15].…”

“…This work stands out from related ones by employing reactive actuation strategies in the dynamic workload of an unknown applications' set. Accurate thermal monitoring supported by a hardware accelerator [14] enables the proposed management strategies. Our management strategies allow more flexibility and adaptability to deal with thermal hotspots by running effective and low-complexity computational algorithms.…”

“…In this work, CM is in charge of mapping and migrating tasks and GM accumulates the application admission to the CM role. This work uses the TEA (Temperature Estimation Accelerator) IP [14]. TEA implements in hardware a simplified version of the MatEx temperature estimation method [17].…”