NVIDIA Scalable Hierarchical Aggregation and Reduction Protocol (SHARP)

Deployment Guide Revision History


Revision

Date

Description

3.11.0

May 08, 2025

  • Added 

3.10.3

February 06, 2025

3.10.2

December 24, 2024

N/A

3.9.0

November 07, 2024

  • Added the following new sections:(3.11.0) Using NVIDIA SHARP with UCC(3.11.0) sharp_am Telemetry(3.11.0) HCA-to-TOR Benchmark Tool

  • Replaced most instances of MLNX_OFED with DOCA-Host throughout the document.

3.8.0

August 14, 2024

3.7.0

May 5, 2024

In , updated the following parameters: 

  • smx_ to smx_sock_addr_family

  • ucx_interface to smx_ucx_interface

3.6.0

February 8, 2023

N/A

3.5.0

November 7, 2023

  • Added 

  • Updated 

3.4.0

August 07, 2023

  • Added UFM Appliance Firewall Settings under 

  • Updated Dynamic vs. Static Allocation Mode

  • Updated Dynamic Trees Allocation Algorithms

  • Added Operating NVIDIA SHARP with PKeys

  • Under :Updated SHARP_COLL_JOB_QUOTA_PAYLOAD_PER_OST flag descriptionUpdated the default value of SHARP_COLL_OSTS_PER_GROUP flagUpdated SHARP_COLL_SAT_LOCK_BATCH_SIZE flag descriptionAdded the default value for SHARP_COLL_NUM_COLL_GROUP_RESOURCE_ALLOC_THRESHOLD flagUpdated the default value for SHARP_COLL_ENABLE_MCAST_TARGET flagUpdated SHARP_COLL_POLL_BATCH descriptionAdded the default value for SHARP_COLL_ERROR_CHECK_INTERVALUpdated SHARP_COLL_JOB_PRIORITY flag descriptionUpdated SHARP_COLL_ENABLE_PCI_RELAXED_ORDERING flag description

3.3.0

May 1, 2023

3.1.1

November 30, 2022

Added 


.

3.1.0

October 31, 2022

2.7.0

May 3, 2022

2.6.1

December 5, 2021

Updated the following sections:


Last updated: