Project Overview#

Caching is critical in data centers to resolve client queries rapidly. The alternative to querying the backend database is a slow process and unsuited to many real-time services today. Unfortunately, the capacity of caches in modern infrastructures is overgrowing the memory capacity existing physical memory technologies can deliver.

This project will explore growing caches over hybrid memory and fast storage devices (e.g., NVMe solid-state drives). We ask whether our approach can satisfy real-time response times while building a huge cache. We will work with Twitter’s Pelikan infrastructure, which is deployed in production at Twitter’s data centers.


  • Experience with programming in C or C++ or Java.
  • Experience with Rust is a plus.


bars search times arrow-up