Randolf Richardson, DevNet SysOp 29 wrote:
>>>There is no need for manually storing files on filesystem, because
>>>large objects are doing that for You. I am storing whole binary files
>>>in blobs(synonym for large objects from some other platforms), and I do
>>>not remember that I had a single problem with that. Do not forget that
>>>libpq has great support for large objects, and you can store large
>>>object without actually storing them on server filesystem, so You do
>>>not need any file permissions on "upload directory" or something like
>>>that.
>>
>> The pictures are prepared for web. Storing in files is faster from
>>that side. That system is calling db every 30 mins...
>
> [sNip]
>
> Has anyone done any benchmarks on this to know what the performance
> differences are for downloads (I'm not concerned about uploads since they're
> far less frequent) from a web server such as Apache HTTPd accessing the file
> system directly vs. a CGI script/program sending it dynamically through
> Apache HTTPd?
CGI will be slow. Of course.
If you write fastCGI, that would be hell lot fast. For ultrafast static content
delivery, you should use small web servers like boa. It outperforms apache2 by
upto 3 times for statis delivery..
Shridhar