[Networkit] networkit algorithms on multiple computing nodes or multiple computers

Meyerhenke, Henning (ITI) meyerhenke at kit.edu
Wed Nov 15 15:51:45 CET 2017


Dear Francesco,

We have analyzed networks with up to ten billion edges on a low-cost
server (256 GB RAM). Getting 1 TB RAM is not that expensive these days,
giving you around 40 billion edges.

For many purposes this is good enough. If not for yours, you will have
to find a different solution indeed, I have to admit.

Just take into account that you either lose a lot of programmer
productivity or execution performance if you leave the shared-mem realm.

Best,
Henning



Am 15.11.17 um 15:08 schrieb Francesco L.:
> OK thank you very much,
> So is not it possible to perform on multiple calculation nodes?
> 
> Actually since we wanted to use the networkit tool to analyze social networks we would have Big Data,
> but if this is not possible, we will find other solutions.
> 
> thank you,
> Francis
> 
>> Il giorno 15 nov 2017, alle ore 14:20, Henning Meyerhenke <henning.meyerhenke at kit.edu> ha scritto:
>>
>> Dear Francesco,
>>
>> So far, NetworKit is designed for shared-memory systems. Dealing with
>> multiple compute nodes is far more involved in terms of programming
>> and/or getting high performance. Plus, for many applications it should
>> not be necessary to scale out of shared memory (mem is cheap...) -- it
>> is the best way to go unless your data is really too big.
>>
>> There are some plans to scale beyond, but do not expect them to become
>> reality any time soon. They depend on demand, funding, resources, and
>> real need of applications -- a very tricky combination.
>>
>> Best,
>> Henning
>>
>>
>>
>>
>> Am 15.11.17 um 13:29 schrieb Francesco landolina:
>>> Good morning everyone,
>>> Thank you very much for the many times that you helped me!
>>> I was wondering whether it would be possible to use the networkit algorithms on multiple computing nodes or multiple computers, so that it could be distributed over multiple machines.
>>>
>>> thank you so much,
>>> Francesco
>>> _______________________________________________
>>> NetworKit mailing list
>>> NetworKit at ira.uni-karlsruhe.de
>>> https://lists.ira.uni-karlsruhe.de/mailman/listinfo/networkit
>>>
>>
>> -- 
>>
>> =======================================
>> Prof. Dr. Henning Meyerhenke
>>
>> Affiliation since Sep 15, 2017:
>> University of Cologne
>>
>> Previous affiliation:
>> Karlsruhe Institute of Technology (KIT)
>> =======================================
>>
> 



More information about the NetworKit mailing list