Heterogeneous Platforms and Systems

Heterogeneous Platforms and Systems are becoming the norm for mobile devices, server systems, and large-scale machines. This project investigates system support to exploit and deal with heterogeneity, at levels of abstraction ranging from middleware, to toolchains, to instrumentation, and operating systems. Concerning individual platforms, our research targets heterogeneous platforms with multicore CPUs and/or integrated and discrete GPUs, and the complex memory systems of future large-memory servers with on-chip fast RAM, die-stacked RAM, NVRAM, and SSD memories. Concerning larger-scale systems, we are considering cloud-hosting datacenters and high end machines like those planned for the exascale era. Current specific research initiatives and students investigating hybrid computing resources include -

  • Efficient execution of irregular or multi-tenant codes on heterogeneous cores - Naila Farooqi, Dipanjan Sengupta
  • Scientific applications on heterogeneous platforms - Alex Merritt, Kavitha Chandrasekar
  • Fine-grained accelerator sharing via JIT injection of resource management constructs - Anshuman Goswami
  • Leveraging NVRAM capacity and persistence properties for next generation workloads - Sudarsun Kannan, Jian Huang
  • The PMFS file system for scalable use of non-volatile memory resources - Dulloor Rao (and others) - Intel Portland

Recent Publications (more on individual student pages) -

  • "Scheduling multi-tenant cloud workloads on accelerator-based systems" Dipanjan Sengupta, Anshuman Goswami, Karsten Schwan, Krishna Pallavi. International Conference for High Performance Computing, Networking, Storage and Analysis, SC14
  • "Slices: Provisioning Heterogeneous HPC Systems" Alexander Merritt, Naila Farooqui, Magdalena Slawinska, Ada Gavrilovska, Karsten Schwan, Vishakha Gupta. Annual Conference on Extreme Science and Engineering Discovery Environment, XSEDE14
  • "HeteroCheckpoint: Efficient Checkpointing for Accelerator-Based Systems" Sudarsun Kannan, Naila Farooqui, Ada Gavrilovska, Karsten Schwan. 44th Annual IEEE/IFIP International Conference on Dependable Systems and Networks, DSN14
  • "Efficient instrumentation of gpgpu applications using information flow analysis and symbolic execution" Naila Farooqui, Karsten Schwan, Sudhakar Yalamanchili. Proceedings of Workshop on General Purpose Processing Using GPUs, GPGPU14

Associated Faculty - Karsten Schwan, Sudhakar Yalamanchili, Ada Gavrilovska, Hyesoon Kim, Jeffrey Young.

Current students - Dulloor Rao, Naila Farooqui, Sudarsun Kannan, Alex Merritt, Dipanjan Sengupta, Jian Huang, Anshuman Goswami, Kavitha Chandrasekar Alumni - Gregory Diamos, Vishakha Gupta, Andrew Kerr, Vishal Gupta, Priyanka Tembey, Min Lee

Older HyVM projects.

Thank you to our sponsors, including Intel Corporation, HP Labs, and Microsoft.