MMMohamed Moataz El Zeininmohamedelzein.hashnode.dev·1d ago · 12 min readHow we optimized our sessions ingestion flow, cutting the latency by 97% while also cutting costs by 95% and no longer getting pagedSome Context Yup, the numbers above are real. For one of our biggest tenants, our CDNs would cross 70~80 million RPM at peaks. They had extremely spiky traffic, where it would go from a comfortable 2000
MMMohamed Moataz El Zeininmohamedelzein.hashnode.dev·Aug 9 · 10 min readHow I managed to build and scale Flink to 300+ million RPM in a stateful exactly-once workflow, and how you could tooWe’ve recently had some huge changes in our entire ingestion flow, and needed to bring on Flink. Our traffic spikes have crossed hundreds of millions of requests per minute and hundreds of terabytes o11A
MMMohamed Moataz El Zeininmohamedelzein.hashnode.dev·Apr 15, 2025 · 15 min readFork, pipes, sockets, and inter-process communicationFork() So, what is fork()? I briefly introduced it in the previous article, but let's get into more details now. So fork() creates an entirely new process by literally copying said process and creating a new one. The child process is essentially an e...00
MMMohamed Moataz El Zeininmohamedelzein.hashnode.dev·Jul 6, 2024 · 17 min readStreams, system calls, file descriptors, and inodes. A small trip through how file access and manipulation work.We all know that system calls are inherently "slow". Regardless of the reasoning behind why system calls are slow, it is common practice to try and "batch" them. A great example of that is how the read() and write() system calls are "batched" or "buf...01K
