Amazon’s Cachee Delivers 31ns Reads for 49KB Post-Quantum Keys
Summarize this article with:
Amazon Web Services is now offering Cachee, a new in-process cache engine designed to address the increasing data demands of post-quantum cryptography. While current systems struggle with the increased size of these new keys, Cachee delivers consistently fast read times of 31 nanoseconds, regardless of value size, from 64-byte tokens to 49,856-byte SLH-DSA-256f signatures. This performance is a critical advancement, as ML-KEM-1024 public keys, at 1,568 bytes, are 10 to 100 times larger than their 32-byte ECDH counterparts. Built in Rust, Cachee achieves 32 million operations per second on a single thread, offering a solution to network latency issues inherent in traditional caching systems like Redis, which requires 0.9 milliseconds for a 17KB SLH-DSA signature. Post-Quantum Key Size Performance: ML-KEM, ML-DSA, SLH-DSA The shift to post-quantum cryptography requires a fundamental rethinking of data handling, as key sizes are expanding dramatically. Signatures generated by the SLH-DSA-256f algorithm, for instance, reach 49,856 bytes, a scale previously unseen in conventional cryptographic systems. This increase in data volume presents a significant challenge for network-bound caches like Redis, where latency traditionally scales linearly with value size, impacting application performance. Cachee, a new in-process cache engine, directly addresses this issue with a design focused on predictive caching for these larger key materials. Performance benchmarks reveal Cachee’s ability to deliver consistent read times, achieving 31 nanoseconds regardless of the data size, even when handling the substantial 49KB SLH-DSA signatures. This contrasts sharply with Redis, which requires 0.9 milliseconds for a 17KB SLH-DSA signature, demonstrating a substantial reduction in latency. Tests show 31 nanosecond reads for 200KB PDFs and 1MB video posters. The need for such optimization is underscored by a direct comparison of key sizes: ML-KEM-1024 public keys require 1,568 bytes, a 10- to 100-fold increase over the 32 bytes used by the established ECDH standard. Cachee’s developers emphasize that value size does not affect read speed, a critical distinction from traditional caching systems. Cachee In-Process Engine: 31ns Reads & Redis Compatibility The increasing demands of post-quantum cryptography are forcing a re-evaluation of caching strategies, as classical systems struggle with dramatically increased key sizes. Where ECDH public keys previously measured 32 bytes, ML-KEM-1024 equivalents now require 1,568 bytes to transmit and process. This expansion extends to digital signatures, with SLH-DSA-256f signatures reaching 49,856 bytes, highlighting the scale of the challenge for network-dependent caches. Traditional approaches, like Redis, experience latency that increases proportionally with value size, creating a bottleneck as data volumes grow. Cachee addresses this issue with an in-process design, aiming to eliminate network overhead and deliver consistent performance regardless of data size. The engine, built in Rust, achieves a reported 31 nanoseconds for reads, even with the 49KB SLH-DSA signatures. Tests demonstrate this claim across a range of data sizes, from 64-byte session tokens to 200KB PDFs, all processed in 31 nanoseconds, and Redis requires 0.9 milliseconds for a 17KB SLH-DSA signature. The core benefit is highlighted by the claim that large-value caching is now solved. 31 nanoseconds regardless of value size. Cachee SaaS Pricing & Deployment: From Free Trial to Unlimited Ops Cachee, a new in-process cache engine, is being deployed through a tiered Software as a Service model, reflecting a strategy to accommodate diverse operational scales as organizations transition to post-quantum cryptography. The company offers a free tier allowing for one million operations as a trial, immediately addressing the need for evaluation in a landscape demanding increasingly complex key management. Beyond the initial trial, Cachee provides a pay-as-you-go option at 0.00002 per cache operation, alongside subscription levels starting at 199 monthly for 20 million operations, scaling up to an “Unlimited” plan at $99,999 per month. This highest tier removes operational caps and includes a 10TB memory allocation. The pricing structure directly responds to the increasing data demands of post-quantum key exchange; for instance, SLH-DSA-256f signatures reach 49,856 bytes, a substantial increase compared to the 64 bytes of Ed25519 signatures used in classical cryptography. Cachee’s architecture is designed to mitigate the performance impact of these larger keys, delivering 31 nanosecond read times regardless of value size, a speed that contrasts sharply with Redis, which requires 0.9 milliseconds for a 17KB SLH-DSA signature. Deployment is streamlined via a simple, single-command installation on Amazon EC2 instances, eliminating the need for lengthy signup processes or complex infrastructure management. Cachee’s developers emphasize that all plans are deployable from the terminal, and billing is handled directly through the terminal via ‘cachee plan upgrade’. Vendor support is available via email with a response time of under 24 hours for all plans, and enterprise customers receive priority support with a dedicated Slack channel, ensuring organizations can rapidly integrate and scale the solution as their post-quantum security needs evolve. Source: https://aws.amazon.com/marketplace/pp/prodview-m23wqhjhybeu6 Tags: