Hi, 
Did you try implementing your filter in a user defined function?
https://docs.arangodb.com/3.1/AQL/Extending/
Its implemented in Javascript - probably not slower than accessing large 
amounts of data via NFS ;-)

If the performance doesn't suit, porting it to C++ and running a patched 
arangod is probably the best way here. 

Please note that filters can make use of indices, and path filters can 
abort traversals during the execution, so performance wise it would 
probably be better to find a way to use them - at least as pre-condition.

Cheers, 
Willi

On Thursday, February 23, 2017 at 8:24:22 PM UTC+1, Alexandre Rostovtsev 
wrote:
>
> Hi all,
>
> With ArangoDB, is it feasible for an external process to open the data 
> files of a live database server in read-only mode, and perform graph 
> traversal using them?
>
> Assuming competent c++ developers and a willingness to read, patch and 
> customize Arango source, but not rewrite half of it.
>
> (We are trying to traverse large graphs with filters that cannot be 
> efficiently implemented on the server side. We can live with the overhead 
> of mmap over nfs, but wish to avoid the overhead of tens of thousands of 
> http requests.)
>
> -Alexandre Rostovtsev.
>

-- 
You received this message because you are subscribed to the Google Groups 
"ArangoDB" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to