ACM Transactions on Computer Systems (TOCS)
Operating systems: design and implementation
Operating systems: design and implementation
Solaris 2.X: internals & architecture
Solaris 2.X: internals & architecture
Forecasting Disk Resource Requirements for a Usenet Server
LISA '93 Proceedings of the 7th USENIX conference on System administration
The Brave Little Toaster Meets Usenet
LISA '96 Proceedings of the 10th USENIX conference on System administration
Scalability in the XFS file system
ATEC '96 Proceedings of the 1996 annual conference on USENIX Annual Technical Conference
Secondary storage management for web proxies
USITS'99 Proceedings of the 2nd conference on USENIX Symposium on Internet Technologies and Systems - Volume 2
Hi-index | 0.00 |
When Usenet News servers were first implemented, the design principle of storing each Usenet article in a separate file appeared to be sound. However, the number of Usenet News articles posted per day has grown phenomenally in the past decade and shows no sign of abating. To stay ahead of the growth curve, Usenet administrators have been forced to buy faster machines, more RAM, and many more disk drives. Many of the performance limitations are caused by interactions with the underlying OS's filesystem, which is usually a Berkeley Fast Filesystem (FFS) derivative. The Cyclic News Filesystem (CNFS) was designed to avoid most of FFS's major problems when used with INN: synchronous file linking/unlinking and sequential scanning of directory files. Articles are stored within a relative handful of large files, either as regular files on top of a standard filesystem or as block disk devices. Articles are stored sequentially within each file, resuming at the beginning of the file when the end is reached. Disk activity is reduced by an order of magnitude.