Bandwidth Engineering Topology 
Engineering at Meta
AUGUST 5, 2024
Topology We built a dedicated backend network specifically for distributed training. To support large language models (LLMs), we expanded the backend network towards the DC-scale, e.g., incorporating topology-awareness into the training job scheduler. We designed a two-stage Clos topology for AI racks, known as an AI Zone.
Engineering at Meta
SEPTEMBER 7, 2023
However, traditional full workload benchmarking presents several challenges: Difficulty in forecasting future system performance : When designing an AI system, engineers frequently face the challenge of predicting the performance of future systems. Our visualization tool can precisely highlight these imbalances, as shown by the below figure.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Kentik
MARCH 13, 2023
Yes, there’s something to say about how applications are written, but on the public internet side, we’ve seen a decrease in latency, cost, and a massive increase in available bandwidth. So what does this mean for today’s enterprise network engineer? This coincided with the advent of the public cloud like AWS, Azure, GCP, etc.
Kentik
APRIL 4, 2023
More than anything, reliability becomes the principal challenge for network engineers working in and with the cloud. Even the most detailed reliability engineering can be easily undermined in an insecure network. While there is much to be said about cloud costs and performance , I want to focus this article primarily on reliability.
Engineering at Meta
JUNE 12, 2024
We optimized the RoCE cluster for quick build time, and the InfiniBand cluster for full-bisection bandwidth. We implemented collective communication patterns with network topology awareness so that they can be less latency-sensitive. The post How Meta trains large language models at scale appeared first on Engineering at Meta.
Engineering at Meta
MARCH 12, 2024
Among other benefits, Hammerspace enables engineers to perform interactive debugging for jobs using thousands of GPUs as code changes are immediately accessible to all nodes within the environment. The post Building Meta’s GenAI Infrastructure appeared first on Engineering at Meta.
High Scalability
FEBRUARY 19, 2021
This means a system that is not merely available but is also engineered with extensive redundant measures to continue to work as its users expect. reliability situations, where continuity of service is essential, with redundant elements continuously in-service, such as with airplane engines. This ensures reliability.
Expert insights. Personalized for you.
132 
Let's personalize your content