Re: Merge algorithms for large numbers of "tapes"

"Luke Lonergan" <llonergan(at)greenplum(dot)com> writes:
> Consider that a popular commercial database, running on a 6-disk RAID5 with
> one filesystem, performs external sorting 4 times faster (1/4 of the time)
> than Postgres using a two pass sort. There is no special optimization of
> the I/O path involved, it's simply a matter of using a modern external
> sorting approach (no tapes).

I think this argumentation hinges on some irrational aversion to the
word "tape". Given adequate work_mem, the CVS-tip behavior is exactly
what you propose already (at least for the cases where we don't need
random access to the sort result). AFAICS the only result of removing
the support for multipass merge is that the code would fail, rather than
run slowly, if it didn't have adequate work_mem for a particular
problem. Somehow I don't see that as an improvement.

regards, tom lane