Projects
Here's a list of current projects running on PRObE resources
ExAlt: Robust Exa-byte Storage
The ultimate goal of this project is to build a storage system which is scalable to exa-bytes and robust against arbitrary faults. Systems with such scale can consist of tens of thousands of machines, each with 10-20 disks, and geo-replicated data centers for durability and availability. We're expecting to provide end-to-end correctness guarantees to this system despite arbitrary faults like disk corruption, memory corruption, or CPU error.
FAWN: A Fast Array of Wimpy Nodes
FAWN is a fast, scalable, and energy-efficient cluster architecture for data-intensive computing. A FAWN cluster links together a large number of "wimpy" nodes built using energy-efficient processors and small amounts of flash memory into an ensemble cluster that can perform the same amount of work as a traditional cluster but at a fraction of the power.
PLFS: Parallel Log Structured Filesystem
The Parallel Log Structured Filesystem Project (PLFS) is a joint project between PDL and Los Alamos Nation Labs (LANL) that is part of the Institute for Reliable High Performance Information Technology (IRHPIT) collaboration between CMU and LANL. PLFS is an open source filesystem that can be inserted between scientific applications and parallel storage system to greatly improve I/O performance.
USRC: Ultrascale Systems Research Center
The Ultrascale system research center is focused on investigating large-scale to exascale distributed system software. PRObE allocations will be used for scaling experiments with our code and traditional code for comparison studies and validation of simulations.
SMR-Filesystem: File systems for Shingled Disks
We use the PRObE hardware for speed experiments of an emulated Shingled Disk filesystem.
VCMTP: A Reliable Message Multicast Transport Protocol for Virtual Circuits
Using the PRObE testbed to explore how a new multicast transport protocol for high-speed scientific data transfers over virtual circuits.
YCSB++: Benchmarking Scalable Table Stores
To test the scalability of YCSB++ and test the effectiveness of different scalable table stores (key-vale, NoSQL databases, etc)