From: William P. <wil...@ya...> - 2010-12-15 15:19:55
|
On Dec 15, 2010, at 3:08 AM, Roderic Page wrote: > However, generating the NEXML file can take what seems like an age, > particularly for large data sets. Yup. Caching would help a lot. But we also need to address what I think is the main cause of this problem: the matrixelement table, that is now probably larger than 200 GB -- so PostgreSQL is not performing well with it. The result is that morphological matrices are very slow to download -- and some are big enough that we get a proxy time-out before the data have been assembled. This is particularly true for matrices with a nucleotide-indel mix, since they are both big and require access to this table. Ideally, any predefined datatypes (ie DNA and protein) should not store records in the matrixelement table -- that would shrink it by several orders of magnitude, with big performance gains. bp |