Towards a Better SCM: Revlog and Mercurial

Large projects need scalable, performant, and robust software configuration management systems. If common revision control operations are not cheap, they present a large barrier to proper software engineering practice. This paper will investigate the theoretical limits on SCM performance, and examines how existing systems fall short of those ideals.

I then describe the Revlog data storage scheme created for the Mercurial SCM. The Revlog scheme allows all common SCM operations to be performed in near-optimal time, while providing excellent compression and robustness.

Finally, I look at how a full distributed SCM (Mercurial) is built on top of the Revlog scheme, some of the pitfalls we've surmounted in on-disk layout and I/O performance and the protocols used to efficiently communicate between repositories.


Download PDF.