The International Conference for High Performance Computing, Networking, Storage and Analysis
The MDTM Project.
Authors: Liang Zhang (Fermi National Accelerator Laboratory), Tan Li (Stony Brook University), Yufei Ren (Stony Brook University), Phil DeMar (Fermi National Accelerator Laboratory), Shudong Jin (Stony Brook University), Dantong Yu (Brookhaven National Laboratory), Wenji Wu (Fermi National Accelerator Laboratory)
Abstract: Multicore and manycore have become the norm for high-performance computing. These new architectures provide advanced features that can be exploited to design and implement a new generation of high-performance data movement tools. DOE ASCR program is funding FNAL and BNL to collaboratively work on a Multicore-Aware Data Transfer Middleware (MDTM) project. MDTM aims to accelerate data movement toolkits on multicore systems. Essentially, MDTM consists of two major components: MDTM middleware services to harness multicore parallelism and make intelligent decision to align CPU, memory, and I/O device together to expedite higher layer applications, and 2) MDTM data transfer applications (client or server) that utilize the middleware to reserve and manage multiple CPU cores, memory, network devices, disk storage as an integrated resource entity and achieve high throughput and improved quality of servers compared with those existing approaches.