Hitoshi Harada <umi(dot)tanuki(at)gmail(dot)com> wrote:
> So, as the first step we implement it by tuplestore with the present
> libpq, but for further improvement we need to refactor or to extend
> our libpq to buffer some sized tuples. Or invent another
> more-data-fetching-oriented protocol like existing copy?
Before starting such optimization, we need some research for which
part is the performance bottleneck when sending large amount of data.
IMHO, since network is typically slower than in-memory data copy,
data compression and burst-transport would be more effective
even if they consume local CPU resources to convert the data format.
Regards,
---
Takahiro Itagaki
NTT Open Source Software Center