Earlier this year, I had the fortune of joining Perplexity AI, where I finally got to use servers with the most powerful configuration—AWS p5 instances equipped with 8 NVIDIA H100 GPUs interconnected via NVSwitch. What excited me even more was the ultra-high-speed 3200 Gbps network between servers. I thought it would be incredibly cool if I could write a program that could utilize this full 3200 Gbps bandwidth!

Recently, I spent a week exploring this, developed a small proof-of-concept program, and managed to utilize 97% of the bandwidth. I found this exploration process quite interesting, and given that there are very limited articles and tutorials online about RDMA, EFA, libfabric, and high-performance networking, I decided to share what I learned during this week. This serves both as a record and as a beginner’s tutorial.

Those familiar with MLSys might ask: Couldn’t this be done with just one line of PyTorch or NCCL code? Indeed, NCCL is very mature in terms of collective communication and is a cornerstone for large language model training and inference. However, I think collective communication has some limitations in other scenarios:

1. Collective communication requires establishing a global communication domain (MPI World). If you need to dynamically add, remove, or replace nodes in the cluster, you need to stop the entire cluster first.
2. Collective communication uses a synchronous communication model. Whether implemented in blocking or non-blocking mode, it poses a significant mental burden for me. I’m more comfortable with an asynchronous communication model like gRPC.
3. Most importantly, isn’t it fun to build your own wheel?

Since my experimental environment is an AWS p5 cluster, some technical details mentioned in this article might only apply to AWS p5 clusters. However, I hope this article can still provide valuable references for other high-performance networking environments.

Because there’s quite a bit of content, I’ve split it into several articles, which you’re welcome to read:

• Harnessing 3200Gbps Network (0): Introduction
• Harnessing 3200Gbps Network (1): RDMA and EFA
• Harnessing 3200Gbps Network (2): High-Performance Network System Design Philosophy
• Harnessing 3200Gbps Network (3): libfabric
• Harnessing 3200Gbps Network (4): Unidirectional SEND and RECV
• Harnessing 3200Gbps Network (5): Bidirectional SEND and RECV
• Harnessing 3200Gbps Network (6): GPUDirect RDMA WRITE
• Harnessing 3200Gbps Network (7): Queuing and Benchmark [97.433 Gbps (97.4%)]
• Harnessing 3200Gbps Network (8): Bus Topology
• Harnessing 3200Gbps Network (9): Using 32 Network Cards [287.089 Gbps (9.0%)]
• Harnessing 3200Gbps Network (10): Pre-benchmark Warmup [293.461 Gbps (9.2%)]
• Harnessing 3200Gbps Network (11): Multi-threading [355.301 Gbps (11.1%)]
• Harnessing 3200Gbps Network (12): CPU Core Pinning [1237.738 Gbps (38.7%)]
• Harnessing 3200Gbps Network (13): State Sharding [1522.567 Gbps (47.6%)]
• Harnessing 3200Gbps Network (14): Batch Posting [2589.488 Gbps (80.9%)]
• Harnessing 3200Gbps Network (15): Lazy Posting [3108.283 Gbps (97.1%)]

As open-source pre-trained Large Language Models (LLMs) become more powerful and permissive, more and more users are incorporating LLMs into their projects. An essential adaptation step is the integration of domain-specific documents into the pre-trained model, known as fine-tuning.

Often, the additional knowledge from domain-specific documents is minuscule compared to what the pre-trained model already knows. In such scenarios, the Low-Rank Adaptation (LoRA) technique proves valuable.

With LoRA, a fine-tuned model adds fewer than 0.1% of parameters to the pre-trained model. In concrete terms, this means a LoRA fine-tuned model increases storage by only 10~200 MB, depending on the configuration. From a computational standpoint, given the marginal increase in parameters compared to the pre-trained model, the additional computational load is relatively small.

Considering the minimal storage addition and computational overhead, I believe there’s potential in developing a multitenancy fine-tuned LLM serving service. This service could host thousands of LoRA models, all sharing the same backbone LLM. With batching, each user request would invoke a distinct fine-tuned model, thereby amortizing storage and computational costs across various models.

In my previous blog post, I delved into the batching effects in LLM serving. In this post, I’ll detail why multitenancy LoRA serving has immense potential.

Machine learning models relying on batching to improve inference throughput, especially for smaller computer vision models such as ResNet and DenseNet. GPT, as well as other large language models (LLMs), is the hottest model these days. Does batching still apply to GPT and LLMs? Let’s find out.

Every time my code crashes, I rely on core dump files to find out where the crash happened. (See my previous post on how to produce and use a core dump file.) My debugging life has been happy thanks to this approach, until this time. When I loaded GDB with the core dump, I’m so disappointed that all the stack trace is about some system library, none about mine.

TLDR: Check this patch.

Let’s start the journey.

The first time I heard Coq was in college. Since then I always wanted to learn Coq but didn’t know how to. Lots ofhypotheses the tutorials are about logic, which doesn’t seem interesting to me.

I took UW CSE505 last quarter. Prof. Zach Tatlock and the TA Talia Ringer made some wonderful homework to help us learn Coq. While in this course, we proved more practical things, like a programming language interpreter and a regular expression matcher. I actually found myself feeling quite good about proving things in Coq (writing specifications is a completely different other thing), at least when doing homework.

If you are teaching yourself Coq, I’d suggest the following resources:

• Homework of UW CSE505
• Formal Reasoning About Programs is the textbook we used in class. It’s a short textbook with big margins and lots of whitespaces. It comes with detailed Coq code for each chapter and a very powerful Coq library frap.
• Software Foundations is the material for you if you are really into verification.

I’m quite happy that I completed Coq in my wish list. But I’m not a huge believer of machine-checked proof or verification. Before I fully forgot Coq, let me write down some beginner-level tricks I learned when doing homework, just in case you might find it useful or I come back to Coq some decades later.