How YouTube scales its massive MySQL databases

Ever wondered how YouTube scales its massive MySQL databases? I was watching NeetCode “How to Design YouTube” video, and one thing completely blew my mind —

𝗚𝗼𝗼𝗴𝗹𝗲 𝘂𝘀𝗲𝘀 𝗠𝘆𝗦𝗤𝗟 𝘁𝗼 𝘀𝘁𝗼𝗿𝗲 𝗺𝗲𝘁𝗮𝗱𝗮𝘁𝗮.

𝘉𝘶𝘵 𝘩𝘦𝘳𝘦’𝘴 𝘵𝘩𝘦 𝘵𝘸𝘪𝘴𝘵 — 𝘵𝘩𝘦𝘺 𝘥𝘰𝘯’𝘵 𝘶𝘴𝘦 𝘱𝘭𝘢𝘪𝘯 𝘔𝘺𝘚𝘘𝘓. 𝘛𝘩𝘦𝘺 𝘶𝘴𝘦 𝘝𝘪𝘵𝘦𝘴𝘴 — 𝘢𝘯 𝘰𝘱𝘦𝘯-𝘴𝘰𝘶𝘳𝘤𝘦, 𝘴𝘩𝘢𝘳𝘥𝘪𝘯𝘨 𝘮𝘪𝘥𝘥𝘭𝘦𝘸𝘢𝘳𝘦 𝘧𝘰𝘳 𝘔𝘺𝘚𝘘𝘓 𝘵𝘩𝘢𝘵 𝘮𝘢𝘬𝘦𝘴 𝘪𝘵 𝘮𝘢𝘴𝘴𝘪𝘷𝘦𝘭𝘺 𝘴𝘤𝘢𝘭𝘢𝘣𝘭𝘦.

Vitess basically acts as a layer between your app and multiple MySQL instances, handling:

• Sharding (splitting data across servers) • Connection pooling • Query routing • Failover and replication

𝗪𝗵𝗮𝘁’𝘀 𝗲𝘃𝗲𝗻 𝗰𝗼𝗼𝗹𝗲𝗿 — 𝗩𝗶𝘁𝗲𝘀𝘀 𝗶𝘀 𝗰𝗹𝗼𝘂𝗱-𝗻𝗮𝘁𝗶𝘃𝗲.

You can deploy it with Kubernetes, and since it doesn’t rely on local storage, your data isn’t tied to any one pod.

Even if a pod shuts down, Vitess ensures durability — as long as one replica remains alive, your committed transactions are safe.

It’s used by giants like YouTube, and the engineers say —

“Once you start sharding, it becomes addictive.” 😄

Despite its scale, Vitess adds just about 0.2 milliseconds of static overhead — incredibly efficient.

https://lnkd.in/gaMSEhR7 Amazing explaination https://lnkd.in/gqY8-5MN

There are now efforts to bring Vitess-like scaling to PostgreSQL too

𝗠𝘂𝗹𝘁𝗶𝗴𝗿𝗲𝘀 → “𝗩𝗶𝘁𝗲𝘀𝘀 𝗳𝗼𝗿 𝗣𝗼𝘀𝘁𝗴𝗿𝗲𝘀” 𝘗𝘭𝘢𝘯𝘦𝘵𝘚𝘤𝘢𝘭𝘦’𝘴 𝘕𝘦𝘬𝘪 → 𝘢 𝘯𝘦𝘸 𝘴𝘩𝘢𝘳𝘥𝘪𝘯𝘨 𝘴𝘰𝘭𝘶𝘵𝘪𝘰𝘯 𝘧𝘰𝘳 𝘗𝘰𝘴𝘵𝘨𝘳𝘦𝘴, 𝘣𝘶𝘪𝘭𝘵 𝘧𝘳𝘰𝘮 𝘵𝘩𝘦 𝘨𝘳𝘰𝘶𝘯𝘥 𝘶𝘱. 𝘋𝘢𝘵𝘢𝘣𝘢𝘴𝘦𝘴 𝘢𝘳𝘦 𝘦𝘷𝘰𝘭𝘷𝘪𝘯𝘨 𝘧𝘢𝘴𝘵 — 𝘢𝘯𝘥 𝘪𝘵’𝘴 𝘸𝘪𝘭𝘥 𝘵𝘰 𝘴𝘦𝘦 𝘩𝘰𝘸 𝘵𝘰𝘰𝘭𝘴 𝘭𝘪𝘬𝘦 𝘝𝘪𝘵𝘦𝘴𝘴 𝘮𝘢𝘬𝘦 “𝘪𝘯𝘧𝘪𝘯𝘪𝘵𝘦 𝘴𝘤𝘢𝘭𝘦” 𝘢 𝘳𝘦𝘢𝘭𝘪𝘵𝘺 𝘰𝘯 𝘵𝘰𝘱 𝘰𝘧 𝘨𝘰𝘰𝘥 𝘰𝘭𝘥 𝘔𝘺𝘚𝘘𝘓.

Comments (0)