Tuning InfiniBand Interconnects Using Congestion Control

By Adam Dorsey

July 26, 2017

InfiniBand is among the most common and well-known cluster interconnect technologies. However, the complexities of an InfiniBand (IB) network can frustrate the most experienced cluster administrators. Maintaining a balanced fabric topology, dealing with underperforming hosts and links, and chasing potential improvements keeps all of us on our toes. Sometimes, though, a little research and experimentation can find unexpected performance and stability gains.

For example, consider a 1,300-node cluster using Intel TrueScale IB for job communication and a Panasas ActiveStor filesystem for storage. Panasas only communicates to clients via Ethernet and not IB, so a group of Mellanox switches act as gateways from the Panasas Ethernet to the TrueScale IB.

Every system has bottlenecks; in our case, the links to and from these IB/Ethernet gateways showed congestion due to the large amount of disk traffic. This adversely affects the whole cluster — jobs can’t get the data they need, and the increased congestion interferes with other IB traffic as well.

Fortunately, InfiniBand provides a congestion control mechanism that can help mitigate the effects of severe congestion on the fabric. We were able to implement this feature to save the expense and trouble of adding additional IB/Ethernet gateways.

What Is InfiniBand Congestion Control?

InfiniBand is intended to be a lossless fabric. IB switches won’t drop packets for flow control unless they absolutely have to, usually in cases of hardware failure or malformed packets. Instead of dropping packets and retransmitting, like Ethernet does, InfiniBand uses a system of credits to perform flow control.

Communication occurs between IB endpoints, which in turn are issued credits based on the amount of buffer space the receiving device has. If the credit cost of the data to be transmitted is less than the credits remaining on the receiving device, the data is transmitted. Otherwise, the transmitting device holds on to the data until the receiving device has sufficient credits free.

This method of flow control works well for normal loads on well-balanced, non-oversubscribed IB fabrics. However, if the fabric is unbalanced or oversubscribed or just heavily loaded, some links may be oversaturated with traffic beyond the ability of the credit mechanism to help.

Congestion can be observed by checking the IB error counters. When an IB device attempts to transmit data but the receiving device cannot receive data due to congestion, the PortXmitWait counter is incremented. If the congestion is so bad that the data cannot be transmitted before the time-to-live on the packet expires, the packet is discarded and the PortXmitDiscards counter is incremented. If you’re seeing high values of PortXmitWait and PortXmitDiscards counters, enabling congestion control may help manage congestion on your InfiniBand fabric.

How Does InfiniBand Congestion Control Work?

When an IB switch detects congestion on a link, it enables a special bit, called the Forward Explicit Congestion Notification (FECN) bit, which informs the destination device that congestion has been detected on the link. When the destination receives a packet marked with the FECN bit, the destination device notifies the sending device of the congestion via a Backwards Explicit Congestion Notification bit (BECN.)

When the source receives the BECN bit notification from the destination, the sending (source) device begins to throttle the amount of data it sends to the destination. The mechanism it uses is the credits system – by reducing the credits available to the destination, the size and rate of the packets are effectively decreased. The sending device may also add a delay between packets to provide the destination device time to catch up on data.

Over time, the source device increases credits for the destination device, gradually increasing the amount of packets sent. If the destination device continues to receive FECN packets from its switch, it again transmits BECN packets to the source device and the throttling is increased again. Without the reception of BECN packets from the destination device, the source device eventually returns to normal packet transmission. This balancing act is managed by congestion control parameters which require tuning for each environment.

After enabling InfiniBand congestion control and proper tuning, we realized a 15 percent improvement in our Panasas file system benchmark testing. PortXmitDiscards counters were completely clear, and PortXmitWait counters were significantly smaller, indicating that congestion control was doing its job.

Given that no additional hardware or other costs were required to achieve these results, a speed increase of 15 percent plus increased stability of the IB fabric was a nice result.

How Can I Enable InfiniBand Congestion Control?

Congestion control must be enabled on all IB devices and hosts, as well as on the IB subnet manager. This process includes turning on congestion control and setting a congestion control key on each device, as well as tuning the congestion control tables and parameters on each host and switch.

After congestion control is enabled on each IB device, the OpenSM configuration file must be modified to tune the subnet manager’s congestion control manager. Please note that mistuned parameters will either wreak havoc on a fabric or be completely ineffectual, so be careful – and do plenty of testing on a safe “test” system. Never attempt this on a live or production system.

Enabling InfiniBand congestion control had an immediate positive effect on our IB fabric. If you are suffering from issues with fabric congestion, enabling congestion control may provide the similar relief for your fabric as well, without the cost of adding additional hardware.

About the Author

Adam Dorsey is a systems administrator and site lead for RedLine Performance Solutions.

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

Nvidia Appoints Andy Grant as EMEA Director of Supercomputing, Higher Education, and AI

March 22, 2024

Nvidia recently appointed Andy Grant as Director, Supercomputing, Higher Education, and AI for Europe, the Middle East, and Africa (EMEA). With over 25 years of high-performance computing (HPC) experience, Grant brings a Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Houston We Have a Solution: Addressing the HPC and Tech Talent Gap

March 15, 2024

Generations of Houstonian teachers, counselors, and parents have either worked in the aerospace industry or know people who do - the prospect of entering the fi Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire