Hama graph api

[ik...@csd.auth.gr]

Hey there,

I am having a strange issue with the Hama Graph API. I have a cluster  
of 2 servers. The thing is that when I run a hama example which  
implements the BSP api, it works fine. When I run an example which  
uses the Graph API, like SSSP, it starts 2 jobs. The first one is just  
one task job (even if I have set more) and it ends after 1 superstep.  
Then immediately it starts a second job which takes the input.

Can you mind checking on that and tell me if you can run it normally?
I have hama 0.6.3.

Thanks

Re: Hama graph api

[Ηλίας Καπουράνης <ik...@csd.auth.gr>]

Cool! And I was thinking that I was screwing something..! Thanks Edward!


Στις 30/10/2013 4:33 μμ, ο/η Edward J. Yoon έγραψε:
> Hi,
>
> GraphJob consists of two jobs. The first is a partitioning job, the second is the Vertex processing job.
>
> Partitioning job divides a input graph into partitions based on the vertex ID. Thus, Graph API allows to send/receive messages among Vertices.
>
> P.S., The default partitioner of GraphJob is HashPartitioner, hash(vertexID) mod N, where N is the number of partitions.
>
> Thanks.
>
> On Oct 30, 2013, at 9:59 PM, ikapoura@csd.auth.gr wrote:
>
>> Hey there,
>>
>> I am having a strange issue with the Hama Graph API. I have a cluster of 2 servers. The thing is that when I run a hama example which implements the BSP api, it works fine. When I run an example which uses the Graph API, like SSSP, it starts 2 jobs. The first one is just one task job (even if I have set more) and it ends after 1 superstep. Then immediately it starts a second job which takes the input.
>>
>> Can you mind checking on that and tell me if you can run it normally?
>> I have hama 0.6.3.
>>
>> Thanks
>>

Re: Hama graph api

["Edward J. Yoon" <ed...@apache.org>]

Hi,

GraphJob consists of two jobs. The first is a partitioning job, the second is the Vertex processing job.

Partitioning job divides a input graph into partitions based on the vertex ID. Thus, Graph API allows to send/receive messages among Vertices. 

P.S., The default partitioner of GraphJob is HashPartitioner, hash(vertexID) mod N, where N is the number of partitions.

Thanks.

On Oct 30, 2013, at 9:59 PM, ikapoura@csd.auth.gr wrote:

> 
> Hey there,
> 
> I am having a strange issue with the Hama Graph API. I have a cluster of 2 servers. The thing is that when I run a hama example which implements the BSP api, it works fine. When I run an example which uses the Graph API, like SSSP, it starts 2 jobs. The first one is just one task job (even if I have set more) and it ends after 1 superstep. Then immediately it starts a second job which takes the input.
> 
> Can you mind checking on that and tell me if you can run it normally?
> I have hama 0.6.3.
> 
> Thanks
>