Twitter's Growth Usage FE throughput Hardware ★ Performance Engineering at Twitter M writes p/s 100K reads p/s LESS Kiji 0.11 MEM ★ ★ ★ Timeline pool ★ Highly normalized K writes p/s M reads p/s Evan Weaver @evan Longlife Ruby heap Row cache LESS ★ Mark barrier set MEM INFO ★ MEM
![Performance Engineering at Twitter](https://cdn-ak-scissors.b.st-hatena.com/image/square/a50c17706b1984de36747b1bfaddea651e53ac04/height=288;version=1;width=512/https%3A%2F%2F0701.static.prezi.com%2Fpreview%2Fv2%2Fnvk6pdxfsp4dvvrkohqexozkdl6jc3sachvcdoaizecfr3dnitcq_3_0.png)