Distributed Run-Time Resource Manager of many-core computing systems

Bill Tsou 5d8323215f Updated README 6 anos atrás
RCCE_V2.0 d80b6a89a3 Added SCC libraries 6 anos atrás
bin 943b78c912 Code restructuring 6 anos atrás
experiments-input 3122f5844f Added init scripts 6 anos atrás
scripts 74b0fc2b90 Added execution scripts 6 anos atrás
src 943b78c912 Code restructuring 6 anos atrás
Makefile d2b76a4bfa Changed a dir in Makefile 6 anos atrás
README.md 5d8323215f Updated README 6 anos atrás
crosscompile.sh d80b6a89a3 Added SCC libraries 6 anos atrás

README.md

The repository contains all the necessary files to execute DRTRM, a Distributed Run-Time Resource Management framework for parallel applications on many-core systems [1]. The target platform of DRTRM, is Intel Single Chip Cloud Computer (SCC), although the design is intended for portability. In addition, it can be compiled to simulate execution of a many-core system, using a process per core on a Linux system. The design parameters and extensions of DRTRM are published at [1], [2], [3].

In order to be executed on Intel SCC the requirement is that Linux is running on every SCC core and RCCEv2.0 has been installed. For a successful compilation, the relevant crosscompile.sh file must be sourced. For a successul compilation for Intel SCC the flags PLATFORM=SCC and API=gory are required. For the rest of the compilation options please contact billtsou AT microlab.ntua.gr.

The folder experiments-input contains all the input data that were used in the experimental evaluation of DRTRM at [1], [2], [3]. They contain application input data, application arrival scenarios, topology parameters of DRTRM and Controller cores topology (See [1]). In addition scripts for the parsing of execution log files are included.

The folder scripts includes scripts for execution different scenarios of DRTRM. For example 'exec_scr_multi_multiple.sh' takes as parameters: (1) the name of the executable file, (2) the path of examined experimental scenario, (3) the name of the examined experimental scenario, (4) the number of controller cores of DRTRM, (5) the type of input application, (6) the workload intensity of incoming applications, (7) the number of incoming applications, (8) the operating frequency of SCC cores, (9) initial cores of the scenario and (10) input application arrival rate.

[1] Tsoutsouras, V., Anagnostopoulos, I., Masouros, D. and Soudris, D., 2018. A Hierarchi- cal Distributed Runtime Resource Management Scheme for NoC-Based Many-Cores. ACM Transactions on Embedded Computing Systems (TECS), 17(3), p.65.

[2] Tsoutsouras, V., Xydis, S. and Soudris, D.J., 2018. Application-Arrival Rate Aware Dis- tributed Run-Time Resource Management for Many-core Computing Platforms. IEEE Transactions on Multi-Scale Computing Systems (TMSCS).

[3] Tsoutsouras, V., Masouros, D., Xydis, S. and Soudris, D., 2017. SoftRM: Self-Organized Fault-Tolerant Resource Management for Failure Detection and Recovery in NoC Based Many-Cores. ACM Transactions on Embedded Computing Systems (TECS), 16(5s), p.144.