Assistant Professor of Computer Science Sang-Woo Jun recently received funding for two new projects, furthering his research into high-performance and high-efficiency systems. He was awarded $500,000 over three years from the National Science Foundation (NSF) for his grant, “MintCloud: An Elastic Multitenant FPGA Cloud.” The MintCloud project aims to design a cloud platform for elastic allocation of field-programmable gate array (FPGA) resources and implement a development and execution environment for distributed FPGA acceleration in such a cloud environment.
“The goal of MintCloud is to create a hardware and software platform for allocation hardware accelerator resources, independently from the host,” says Jun, “and to provide a flexible programming environment for such elastically allocated FPGAs.” The platform would thus help researchers in industry and academia rapidly develop and deploy complex, high-performance low-power applications at a low cost in the cloud.
Jun was also gifted $75,000 from VMware Research for his project, “Flash Management for Smart SSDs.” This project involves designing and evaluating various applications that could benefit from smart solid-state drives, where computation units are embedded in fast peripheral component interconnect express (PCIe) SSD storage devices.
“One system architecture paradigm gaining traction recently is near-storage processing, where a processing unit is placed on a storage device and performs computation on data as it is read out from storage,” explains Jun. The main benefits of this architecture are that it can save storage bandwidth, reduce latency, and take advantage of the whole internal bandwidth instead of the limited performance allowed by the connection (such as PCIe).
“The collaboration project with VMWare,” says Jun, “is to design and evaluate such an architecture on a number of applications that are important in a cloud/datacenter context, including log analytics and database operations, to measure the end-to-end benefits for an end user.”
When asked about the motivation behind his investigation of unconventional system architectures, Jun points to UC Berkeley Professor David Patterson’s assertion in an IEEE Spectrum article last year. “We are now a factor of 15 behind where we should be if Moore’s Law were still operative,” says Patterson. “Revolutionary new hardware architectures and new software languages, tailored to dealing with specific kinds of computing problems, are just waiting to be developed.”
Jun is doing his part to help develop these revolutionary solutions.
— Shani Murray