Best of Edge Computing — July 2024

1. 1
   Article

   Community Picks·2y

   HTTP vs. WebSockets: Which protocol for your Postgres queries at the Edge

   SQL-over-HTTP has been added to the Neon driver, which previously only supported WebSockets, for querying Postgres databases closer to end-users. WebSockets excel in maintaining consistent connections, offering low latencies for sustained queries but perform slower for single-shot queries. By contrast, HTTP is quicker for single-shot queries but lacks features like session support and interactive transactions. To strike a balance, connection caching was introduced, speeding up HTTP queries by around 10ms. The choice between HTTP and WebSockets depends on query type, user location, and specific APIs used.

   33
2. 2
   Article

   Community Picks·2y

   An easy intro to edge computing

   Edge computing enhances website performance by processing data closer to the user, reducing latency and improving user experience. It involves using networks of servers (CDNs) to cache and serve content more efficiently. Modern edge networks can run code close to users, enabling personalized content, A/B testing, and better error handling. Technologies like Fastly and Glitch aim to make these advanced capabilities accessible to more developers, paving the way for innovative web applications.

   19

   1
3. 3
   Article

   Medium·2y

   The Ultimate Handbook for LLM Quantization

   Large Language Models (LLMs) often require substantial computational resources, making them challenging to run on devices without powerful GPUs. Quantization is a technique that reduces the memory footprint and computational requirements by converting higher-precision weights to lower-precision formats, such as FP32 to INT8. This post delves into various quantization methods, including Post-Training Quantization (PTQ) and Quantization-Aware Training (QAT), and reviews state-of-the-art techniques like LLM.int8(), GPTQ, and QLoRA. These methods help enable LLM deployment on edge devices without significant performance loss.

   13