Issue No. 02 - June (1989 vol. 1)
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/69.87963
<p>The authors present a brief description of data on directory reference patterns collected from a 4.2BSD UNIX system. These data are used to examine the importance of the name lookup overhead involved in opening and using files. The analysis shows that paths in the environment are relatively long and that, in the absence of caching, name resolution overhead accounts for over 70% of the disk blocks referenced to open and use files. These results confirm recent conjectures on the high level of directory activity in UNIX file systems. Directory references exhibit strong locality, though, making caches an effective way to decrease directory overhead. Simulations of a least recently used (IRU) whole directory cache show that a cache holding just ten nodes achieves an 85% hit ratio. The implications of these results on the design of both local and distributed file systems are discussed.</p>
file opening; file use; local file systems; hierarchical file systems; directory reference patterns; 4.2BSD UNIX system; name lookup overhead; paths; environment; name resolution overhead; disk blocks; locality; caches; least recently used; IRU; nodes; distributed file systems; data handling; distributed databases; file organisation
R. Floyd and C. Schlatter Ellis, "Directory Reference Patterns in Hierarchical File Systems," in IEEE Transactions on Knowledge & Data Engineering, vol. 1, no. , pp. 238-247, 1989.