/ucx

Unified Communication X (mailing list - https://elist.ornl.gov/mailman/listinfo/ucx-group)

Primary LanguageCOtherNOASSERTION


follow on Twitter Documentation Status

Unified Communication X

Unified Communication X (UCX) is an award winning, optimized production proven-communication framework for modern, high-bandwidth and low-latency networks.

UCX exposes a set of abstract communication primitives that utilize the best of available hardware resources and offloads. These include RDMA (InfiniBand and RoCE), TCP, GPUs, shared memory, and network atomic operations.

Please visit our documentation site for more details.

Please review our "Membership Voluntary Consensus Standard" and "Export Compliant Contribution Submissions" policies.



Using UCX

Release Builds

Building UCX is typically a combination of running "configure" and "make". Execute the following commands to install the UCX system from within the directory at the top of the tree:

$ ./autogen.sh
$ ./contrib/configure-release --prefix=/where/to/install
$ make -j8
$ make install

NOTE: Compiling support for various networks or other specific hardware may require additional command line flags when running configure.

Developer Builds

$ ./autogen.sh
$ ./contrib/configure-devel --prefix=$PWD/install-debug

*** NOTE: Developer builds of UCX typically include a large performance penalty at run-time because of extra debugging code.

Build RPM package

$ contrib/buildrpm.sh -s -b

Build DEB package

$ dpkg-buildpackage -us -uc

Build Doxygen documentation

$ make docs

OpenMPI and OpenSHMEM installation with UCX

Wiki page

MPICH installation with UCX

Wiki page

UCX Performance Test

Start server:

$ ./src/tools/perf/ucx_perftest -c 0

Connect client:

$ ./src/tools/perf/ucx_perftest <server-hostname> -t tag_lat -c 1

NOTE the -c flag sets CPU affinity. If running both >commands on same host, make sure you set the affinity to different CPU cores.

Running internal unit tests

$ make -C test/gtest test

Known issues

  • UCX version 1.8.0 has a bug that may cause data corruption when TCP transport is used in conjunction with shared memory transport. It is advised to upgrade to UCX version 1.9.0 and above. UCX versions released before 1.8.0 don't have this bug.

  • UCX may hang with glibc versions 2.25-2.29 due to known bugs in the pthread_rwlock functions. When such hangs occur, one of the UCX threads gets stuck in pthread_rwlock_rdlock (which is called by ucs_rcache_get), even though no other thread holds the lock. A related issue is reported in glibc Bug 23844. If this issue occurs, it is advised to use glibc version provided with your OS distribution or build glibc from source using versions less than 2.25 or greater than 2.29.

  • Due to compatibility flaw when using UCX with rdma-core v22 setting UCX_DC_MLX5_RX_INLINE=0 is unsupported and will make DC transport unavailable. This issue is fixed in rdma-core v24 and backported to rdma-core-22.4-2.el7 rpm. See ucx issue 5749 for more details.


Architecture

Component Role Description
UCP Protocol Implements high-level abstractions such as tag-matching, streams, connection negotiation and establishment, multi-rail, and handling different memory types
UCT Transport Implements low-level communication primitives such as active messages, remote memory access, and atomic operations
UCS Services A collection of data structures, algorithms, and system utilities for common use
UCM Memory Intercepts memory allocation and release events, used by the memory registration cache

Supported Transports


Supported CPU Architectures


Licenses

UCX is licensed as:


Our Community


Contributor Agreement and Guidelines

In order to contribute to UCX, please sign up with an appropriate Contributor Agreement.

All contributors have to comply with "Membership Voluntary Consensus Standard" and "Export Compliant Contribution Submissions" policies.

Follow these instructions when submitting contributions and changes.

Publications

To reference UCX in a publication, please use the following entry:

@inproceedings{shamis2015ucx,
  title={UCX: an open source framework for HPC network APIs and beyond},
  author={Shamis, Pavel and Venkata, Manjunath Gorentla and Lopez, M Graham and Baker, Matthew B and Hernandez, Oscar and Itigin, Yossi and Dubman, Mike and Shainer, Gilad and Graham, Richard L and Liss, Liran and others},
  booktitle={2015 IEEE 23rd Annual Symposium on High-Performance Interconnects},
  pages={40--43},
  year={2015},
  organization={IEEE}
}

To reference the UCX website:

@misc{openucx-website,
    title = {{The Unified Communication X Library}},
    key = {{{The Unified Communication X Library}},
    howpublished = {{\url{http://www.openucx.org}}}
}