[EP-tech] Re: Injecting gigabyte-scale files into EPrints archive - impossible?

[hess at ub.uni-heidelberg.de (Florian Heß)]

Am 18.08.2014 um 16:17 schrieb Paolo Tealdi:
> Il 04/08/2014 10:13, Yuri ha scritto:
>> The only option seems to enlarge the tmp :-)
>>
>
> or trying to convince the system to change the TMPDIR to a bigger filesystem.
>

Thank you both, Yuri and Paolo. :-)

I am afraid, though, that this is not quite a sustainable, scalable 
approach. If for instance one used Recollect Plugin or whatever, anyway 
EPrints as a research data store, even bigger files would be to serve, 
say hundred gigabyte large database dumps. EPrints could not handle that 
giant files this way, as it would need absurd *several hundred* gigabyte 
of RAM and swap, let alone other involved bottle-necks in the program 
flow. If you don't mind my assuming that in regard to its easygoing use 
of join("", <STDIN>).

Concerning large files, we thought we could serve files up to a maximum 
of a DVD image. For the time being I will prefer the described copy & 
hack database method (but as said, I'll try use the API next time).

Look, as a public university library it is not overwhelming hardware 
resources which we have got for the given purpose, so we must use wise 
and economically what is available.


Kind regards
Florian

> Best regards,
> Paolo Tealdi
>


-- 
UB Heidelberg (Altstadt)
Pl?ck 107-109, 69117 HD
Abt. Informationstechnik
http://www.ub.uni-heidelberg.de/