On-the-Fly TCP Acceleration with Miniproxy
📜 Abstract
TCP proxies are basic building blocks for many advanced middleboxes. In this paper we present Miniproxy, a TCP proxy built on top of a specialized minimalistic cloud operating system. Miniproxy’s connection handling performance is comparable to that of full-fledged GNU/Linux TCP proxy implementations, but its minimalistic footprint enables new use cases. Specifically, Miniproxy requires as little as 6 MB to run and boots in tens of milliseconds, enabling massive consolidation, on-the-fly instantiation and edge cloud computing scenarios. We demonstrate the benefits of Miniproxy by implementing and evaluating a TCP acceleration use case.
✨ Summary
Summary and subsequent influence
The paper presents Miniproxy, a Xen-based unikernel TCP proxy implemented using MiniOS and a modified lwIP stack. Its design targets dynamically deployed TCP acceleration: proxies split an end-to-end TCP connection into shorter segments, use Early SYN Forwarding to overlap connection establishment, and allow the TCP slow-start phases of the split connections to proceed in parallel. The implementation is event-driven, approximately 600 lines of code, requires at least 6 MB of RAM, and boots in 12 ms under the reported evaluation conditions. In controlled experiments, Miniproxy achieved approximately 1.534 Gb/s on a single TCP flow, compared with 1.462 Gb/s for Varnish configured in the same test environment but using substantially more memory. With a 100 ms client-server RTT, one, two, and three evenly placed proxies reduced Time To First Byte by 25%, 33.3%, and 37.5%, respectively; measured total-transfer-time reductions for the evaluated flows reached 33%, 44%, and 49%.
The clearest documented follow-on use is MOSTO, a dynamically managed overlay for accelerating dynamic content delivery. The later work explicitly states that it implemented its fast-booting TCP proxies using Miniproxy and extended the approach with proxy-chain selection, reactive instantiation, cloud deployment, and switch-assisted offloading. MOSTO was subsequently developed into a peer-reviewed IEEE Journal on Selected Areas in Communications article. (arxiv.org)
The original work was also published in the ACM HotMiddlebox 2016 proceedings, where it appears as a six-author paper on pages 44–49. (dblp.org)
Overall, the paper’s documented influence is primarily methodological and architectural: it provided a lightweight, rapidly instantiated TCP-proxy building block that was directly incorporated into later research on virtualized content delivery and dynamic edge-network services. The available evidence supports research reuse; it does not establish broad commercial deployment.