While these fixes moved the needle, they had limitations. By analyzing our database traffic, we learned that writes— like gathering, updating, or deleting data—contributed to a significant portion of database utilization. Additionally, not all reads or data fetching could be moved to replicas due to application sensitivity to replication lag. So, from both a read and write perspective, we still ne
![The growing pains of database architecture | Figma Blog](https://cdn-ak-scissors.b.st-hatena.com/image/square/f951c3cf7946bb5a73016c02e9948efa6b1d1988/height=288;version=1;width=512/https%3A%2F%2Fcdn.sanity.io%2Fimages%2F599r6htc%2Fregionalized%2F5c8ad1394a1cf21fc39adf7ca3d9f317c8b98b80-2120x1000.png%3Fw%3D1200%26q%3D70%26fit%3Dmax%26auto%3Dformat)