Memory-Centric Computing
Some applications have a working set that is larger than what fits in the main memory of today's computer systems. Systems that have enough memory to keep the entire working set of such applications in memory are called memory-centric systems.[1][2]
Overview[edit]
A memory-centric system consists of multiple independent compute nodes that are connected through a firewall to a shared byte-addressable memory. Each compute node has one or more processors, has local memory and can access both the local memory and the shared memory with regular load and store instructions.[3][4] Optical connections are used between the compute nodes and the shared memory to provide fast access. The firewall determines the access rights to the shared memory of the local node. Processors on a single node access the local memory in a cache coherent way. Accesses to the shared memory from different nodes are not coherent.[5] A challenge with this approach is that accesses to the shared memory are slower than accesses to local memory and hence result in reduced application performance.[6] [7][8]
References[edit]
- ↑ Faraboschi, Paolo; Keeton, Kimberly; Marsland, Tim; Milojicic, Dejan (2015). Beyond Processor-centric Operating Systems (PDF). HotOS.
- ↑ Lohr, Steve (October 31, 2011). "Big Data, Speed and the Future of Computing". New York Times. Retrieved July 22, 2018.
- ↑ Packard, Keith (2015). "The Machine Architecture". Retrieved July 11, 2018.
- ↑ Keeton, Kimberly (2015). The machine: An architecture for memory-centric computing (PDF). In Workshop on Runtime and Operating Systems for Supercomputers (ROSS). Retrieved July 13, 2018.
- ↑ Edge, Jake (August 26, 2015). "A look at The Machine". Retrieved July 11, 2018.
- ↑ Zhu, Guoliang; Lu, Kai; Wang, Xiaoping; Zhou, Xu; Shi, Zhan (2017). "Building emulation framework for non-volatile memory". IEEE Access. 5: 21574–21584.
- ↑ Volos, Haris; Magalhaes, Guilherme; Cherkasova, Ludmila; Li, Jun (2015). Quartz: A lightweight performance emulator for persistent memory software (PDF). 16th Annual Middleware Conference. ACM. pp. 37–49.
- ↑ Harris, Robin (24 March 2017). "What would a memory-centric system look like?". ZDNet. Retrieved 22 September 2018.
This article "Memory-Centric Computing" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:Memory-Centric Computing. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
