Portable SHMEMCache: A High-Performance Key-Value Store on OpenSHMEM and MPI

被引:0
作者
Fu, Huansong [1 ]
Venkata, Manjunath Gorentla [2 ]
Imam, Neena [2 ]
Yu, Weikuan [1 ]
机构
[1] Florida State Univ, Tallahassee, FL 32306 USA
[2] Oak Ridge Natl Lab, Oak Ridge, TN USA
来源
OPENSHMEM AND RELATED TECHNOLOGIES: BIG COMPUTE AND BIG DATA CONVERGENCE, OPENSHMEM 2017 | 2018年 / 10679卷
基金
美国国家科学基金会;
关键词
IMPLEMENTATION; RDMA;
D O I
10.1007/978-3-319-73814-7_8
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
The integration of Big Data frameworks and HPC capabilities has drawn enormous interests in recent years. SHMEMCache is a distributed key-value store built on the OpenSHMEM global address space. It has solved several practical issues in leveraging OpenSHMEM's one-sided operations for a distributed key-value store and providing efficient key-value operations on both commodity machines and supercomputers. However, being based solely on OpenSHMEM, SHMEMCache cannot leverage one-sided operations from a variety of software packages. This results in several limitations for SHMEMCache. First, we cannot make SHMEMCache available to a wider range of platforms. Second, an opportunity for potential performance improvement is missed. Third, there is a lack of deep understanding about how different one-sided operations can fit in with SHMEMCache and other distributed key-values in general. For example, the one-sided operations in OpenSHMEM and MPI have many differences in their interfaces, memory semantics and synchronization methods, all of which can have distinct implications and also increase the complexity in supporting both OpenSHMEM and MPI for SHMEMCache. Therefore, we have taken on an effort on leveraging different one-sided operations for SHMEMCache and proposed a design of portable SHMEMCache. Based on this new framework, we have supported both OpenSHMEM and MPI for SHMEMCache. We have also conducted an extensive set of experiments to compare the performance of the two versions on both commodity machines and the Titan supercomputer.
引用
收藏
页码:114 / 129
页数:16
相关论文
共 30 条
[1]  
Aniszczyk C., 2012, Caching with twemcache
[2]  
[Anonymous], 2005, UPC language specifications v1. 2
[3]  
[Anonymous], 2013, 10 USENIX S NETW SYS
[4]  
[Anonymous], 2010, P 1 ACM S CLOUD COMP, DOI DOI 10.1145/1807128.1807152
[5]  
[Anonymous], 2014, P 11 USENIX C NETW S
[6]  
Appavoo Jonathan., 2010, Proceedings of the 19th ACM International Symposium on High Performance Distributed Computing, P385, DOI [10.1145/1851476.1851534, DOI 10.1145/1851476.1851534]
[7]  
Chapman B., 2010, P 4 C PART GLOB ADDR, P1
[8]  
Chen YZ, 2016, AER ADV ENG RES, V53, P26
[9]   An implementation and evaluation of the MPI 3.0 one-sided communication interface [J].
Dinan, James ;
Balaji, Pavan ;
Buntinas, Darius ;
Goodell, David ;
Gropp, William ;
Thakur, Rajeev .
CONCURRENCY AND COMPUTATION-PRACTICE & EXPERIENCE, 2016, 28 (17) :4385-4404
[10]   No compromises: distributed transactions with consistency, availability, and performance [J].
Dragojevic, Aleksandar ;
Narayanan, Dushyanth ;
Nightingale, Edmund B. ;
Renzelmann, Matthew ;
Shamis, Alex ;
Badam, Anirudh ;
Castro, Miguel .
SOSP'15: PROCEEDINGS OF THE TWENTY-FIFTH ACM SYMPOSIUM ON OPERATING SYSTEMS PRINCIPLES, 2015, :54-70