Re: Suggestions on problem sizes for giraph performance benchmarking

[Amani Alonazi]

Hi Benjamin,

I'm really interesting in this kind of algorithm, if you don't mind to
share it with me. I'm working in Giraph and other pregel - clone system to
demonstrate some graph algorithms such as connected components (strong and
week), max clique, Eulerian Path and others. It would be helpful if you
share with me the code.

Thank you,

On Mon, Jul 9, 2012 at 3:44 PM, Benjamin Heitmann <
benjamin.heitm...@deri.org> wrote:

>
> Hello Stephen,
>
> sorry for the very late reply.
>
> On 28 Jun 2012, at 02:50, Fleischman, Stephen (ISS SCI - Plano TX) wrote:
>
>  Hello Avery and all:
>
> I have a cluster of 10  two-processor/48 GB RAM servers, upon which we are
> conducting Hadoop performance characterization tests.  I plan to use the
> Giraph pagerank and simple shortest path example tests as part of this
> exercise and would appreciate guidance on problem sizes for both tests.
> I’m looking at paring down an obfuscated Twitter dataset and it would save
> a lot of time if someone has some knowledge on roughly how the time and
> memory scales with number of nodes in a graph.
>
> 
>
>
>
> I can provide some suggestions for the kind of algorithm and data which
> does currently surpass the scalability of giraph.
>
> While the limits to my knowledge of Giraph and Hadoop are probably also to
> blame for this, please see the recent discussions on this list,
> and on JIRA for other indications that the scalability of Giraph needs
> improvement:
> * post  by Yuanyuan Tian in the thread "wierd communication errors" on
> user@giraph.apache.org
> * GIRAPH-234 about GC overhead
> https://issues.apache.org/jira/browse/GIRAPH-234?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
>
> If you want to stretch the limits of Giraph, then you need to try an
> algorithm which is conceptually different from PageRank, and you need a big
> data set.
> If you use an algorithm which has complex application logic (maybe even
> domain specific logic), which needs to be embedded in the algorithm,
> then the nodes need to have a lot of state. In addition, such algorithms
> probably send around a lot of messages, and each of the messages might have
> a payload
> which is more complex then one floating point number. In addition, it
> helps to have a graph format, which requires strings on the edges and
> vertices.
> The strings are required for the domain specific business logic which the
> graph algorithm needs to follow.
>
> Finally, imagine a data set which has a big loading time, and where one
> run of the algorithm only provides results for one user.
> The standard Hadoop paradigm is to throw away the graph after loading it.
> So if you have 100s or 1000s of users, then you need a way to execute the
> algorithm multiple times in parallel.
> Again this will add a lot of state, as each of the vertices will need to
> hold one state object for each user who has visited the vertex.
>
> In my specific case, I had the following data and algorithm:
> Data:
> * an RDF graph with 10 million vertices and 40 million edges
> I used my own import code to map the RDF graph to a undirected graph with
> a limit of one edge between any two nodes (so it was not a multi-graph)
> * each vertex and each edge uses a string as an identity to represent a
> URI in the RDF graph (required for the business logic in the algorithm)
>
> Algorithm:
> * spreading activation.
> You can think of it as depth first search guided by domain specific logic.
> A short introduction here:
> https://en.wikipedia.org/wiki/Spreading_activation
> The wikipedia article only mentions using spreading activation on weighted
> graphs, however I used it on graphs which have additional types on the
> edges.
> The whole area of using the semantics of the edges to guide the algorithm
> is an active research topic, so thats why I can't point you to a good
> article on that.
> * parallel execution:
> I need to run the algorithm once for every user in the system, however
> loading the data set takes around 15 minutes alone.
> So each node has an array of states, one for each user for which the
> algorithm has visited a node.
> I experimented with user numbers between 30 and 1000, anything more did
> not work for concurrent execution of the algorithm.
>
> Infrastructure:
> * a single server with 24 Intel Xeon 2.4 GHz cpus and 96 GB of RAM
> * Hadoop 1.0, pseudo-distributed setup
> * between 10 and 20 Giraph workers
>
>
> A few weeks ago I stopped work on my Giraph based implementation, as
> Giraph ran out of memory almost immediately after loading and initialising
> the data.
> I made sure that the Giraph workers do not run out of memory, so it was
> probably due to IPC and messaging.
> The general discussion on the Giraph mailing list strongly indicates that
> I did hit the current IPC scalability limits.
>
> Currently I am working on a non-Hadoop version of the algorithm which is
> not as scalable but which is fast for *one* user. ( less then a second per
> user, but 

Re: Suggestions on problem sizes for giraph performance benchmarking

[Avery Ching]

You should try using the appropriate memory settings (i.e.  
-Dmapred.child.java.opts="-Xms30g -Xmx30g -Xss128k") for a 30 GB heap.  
This depends on how much memory you can get.


Avery

On 7/9/12 5:57 AM, Amani Alonazi wrote:
Actually, I had the same problem of running out of memory with Giraph 
when trying to implement strongly connected components algorithm on 
Giraph. My input graph is 1 million nodes and 7 million edges.


I'm using cluster of 21 computers.


On Mon, Jul 9, 2012 at 3:44 PM, Benjamin Heitmann 
mailto:benjamin.heitm...@deri.org>> wrote:



Hello Stephen,

sorry for the very late reply.

On 28 Jun 2012, at 02:50, Fleischman, Stephen (ISS SCI - Plano TX)
wrote:


Hello Avery and all:

I have a cluster of 10  two-processor/48 GB RAM servers, upon
which we are conducting Hadoop performance characterization
tests.  I plan to use the Giraph pagerank and simple shortest
path example tests as part of this exercise and would appreciate
guidance on problem sizes for both tests.  I’m looking at paring
down an obfuscated Twitter dataset and it would save a lot of
time if someone has some knowledge on roughly how the time and
memory scales with number of nodes in a graph.




I can provide some suggestions for the kind of algorithm and data
which does currently surpass the scalability of giraph.

While the limits to my knowledge of Giraph and Hadoop are probably
also to blame for this, please see the recent discussions on this
list,
and on JIRA for other indications that the scalability of Giraph
needs improvement:
* post  by Yuanyuan Tian in the thread "wierd communication
errors" on user@giraph.apache.org 
* GIRAPH-234 about GC overhead

https://issues.apache.org/jira/browse/GIRAPH-234?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

If you want to stretch the limits of Giraph, then you need to try
an algorithm which is conceptually different from PageRank, and
you need a big data set.
If you use an algorithm which has complex application logic (maybe
even domain specific logic), which needs to be embedded in the
algorithm,
then the nodes need to have a lot of state. In addition, such
algorithms probably send around a lot of messages, and each of the
messages might have a payload
which is more complex then one floating point number. In addition,
it helps to have a graph format, which requires strings on the
edges and vertices.
The strings are required for the domain specific business logic
which the graph algorithm needs to follow.

Finally, imagine a data set which has a big loading time, and
where one run of the algorithm only provides results for one user.
The standard Hadoop paradigm is to throw away the graph after
loading it.
So if you have 100s or 1000s of users, then you need a way to
execute the algorithm multiple times in parallel.
Again this will add a lot of state, as each of the vertices will
need to hold one state object for each user who has visited the
vertex.

In my specific case, I had the following data and algorithm:
Data:
* an RDF graph with 10 million vertices and 40 million edges
I used my own import code to map the RDF graph to a undirected
graph with a limit of one edge between any two nodes (so it was
not a multi-graph)
* each vertex and each edge uses a string as an identity to
represent a URI in the RDF graph (required for the business logic
in the algorithm)

Algorithm:
* spreading activation.
You can think of it as depth first search guided by domain
specific logic.
A short introduction here:
https://en.wikipedia.org/wiki/Spreading_activation
The wikipedia article only mentions using spreading activation on
weighted graphs, however I used it on graphs which have additional
types on the edges.
The whole area of using the semantics of the edges to guide the
algorithm is an active research topic, so thats why I can't point
you to a good article on that.
* parallel execution:
I need to run the algorithm once for every user in the system,
however loading the data set takes around 15 minutes alone.
So each node has an array of states, one for each user for which
the algorithm has visited a node.
I experimented with user numbers between 30 and 1000, anything
more did not work for concurrent execution of the algorithm.

Infrastructure:
* a single server with 24 Intel Xeon 2.4 GHz cpus and 96 GB of RAM
* Hadoop 1.0, pseudo-distributed setup
* between 10 and 20 Giraph workers


A few weeks ago I stopped work on my Giraph based implementation,
as Giraph ran out of memory almost immediately after loading and
initialising the data.
I made sure that the Giraph workers do not run out of memory, so
it was

Re: Suggestions on problem sizes for giraph performance benchmarking

[Amani Alonazi]

Actually, I had the same problem of running out of memory with Giraph when
trying to implement strongly connected components algorithm on Giraph. My
input graph is 1 million nodes and 7 million edges.

I'm using cluster of 21 computers.


On Mon, Jul 9, 2012 at 3:44 PM, Benjamin Heitmann <
benjamin.heitm...@deri.org> wrote:

>
> Hello Stephen,
>
> sorry for the very late reply.
>
> On 28 Jun 2012, at 02:50, Fleischman, Stephen (ISS SCI - Plano TX) wrote:
>
>  Hello Avery and all:
>
> I have a cluster of 10  two-processor/48 GB RAM servers, upon which we are
> conducting Hadoop performance characterization tests.  I plan to use the
> Giraph pagerank and simple shortest path example tests as part of this
> exercise and would appreciate guidance on problem sizes for both tests.
> I’m looking at paring down an obfuscated Twitter dataset and it would save
> a lot of time if someone has some knowledge on roughly how the time and
> memory scales with number of nodes in a graph.
>
> 
>
>
>
> I can provide some suggestions for the kind of algorithm and data which
> does currently surpass the scalability of giraph.
>
> While the limits to my knowledge of Giraph and Hadoop are probably also to
> blame for this, please see the recent discussions on this list,
> and on JIRA for other indications that the scalability of Giraph needs
> improvement:
> * post  by Yuanyuan Tian in the thread "wierd communication errors" on
> user@giraph.apache.org
> * GIRAPH-234 about GC overhead
> https://issues.apache.org/jira/browse/GIRAPH-234?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
>
> If you want to stretch the limits of Giraph, then you need to try an
> algorithm which is conceptually different from PageRank, and you need a big
> data set.
> If you use an algorithm which has complex application logic (maybe even
> domain specific logic), which needs to be embedded in the algorithm,
> then the nodes need to have a lot of state. In addition, such algorithms
> probably send around a lot of messages, and each of the messages might have
> a payload
> which is more complex then one floating point number. In addition, it
> helps to have a graph format, which requires strings on the edges and
> vertices.
> The strings are required for the domain specific business logic which the
> graph algorithm needs to follow.
>
> Finally, imagine a data set which has a big loading time, and where one
> run of the algorithm only provides results for one user.
> The standard Hadoop paradigm is to throw away the graph after loading it.
> So if you have 100s or 1000s of users, then you need a way to execute the
> algorithm multiple times in parallel.
> Again this will add a lot of state, as each of the vertices will need to
> hold one state object for each user who has visited the vertex.
>
> In my specific case, I had the following data and algorithm:
> Data:
> * an RDF graph with 10 million vertices and 40 million edges
> I used my own import code to map the RDF graph to a undirected graph with
> a limit of one edge between any two nodes (so it was not a multi-graph)
> * each vertex and each edge uses a string as an identity to represent a
> URI in the RDF graph (required for the business logic in the algorithm)
>
> Algorithm:
> * spreading activation.
> You can think of it as depth first search guided by domain specific logic.
> A short introduction here:
> https://en.wikipedia.org/wiki/Spreading_activation
> The wikipedia article only mentions using spreading activation on weighted
> graphs, however I used it on graphs which have additional types on the
> edges.
> The whole area of using the semantics of the edges to guide the algorithm
> is an active research topic, so thats why I can't point you to a good
> article on that.
> * parallel execution:
> I need to run the algorithm once for every user in the system, however
> loading the data set takes around 15 minutes alone.
> So each node has an array of states, one for each user for which the
> algorithm has visited a node.
> I experimented with user numbers between 30 and 1000, anything more did
> not work for concurrent execution of the algorithm.
>
> Infrastructure:
> * a single server with 24 Intel Xeon 2.4 GHz cpus and 96 GB of RAM
> * Hadoop 1.0, pseudo-distributed setup
> * between 10 and 20 Giraph workers
>
>
> A few weeks ago I stopped work on my Giraph based implementation, as
> Giraph ran out of memory almost immediately after loading and initialising
> the data.
> I made sure that the Giraph workers do not run out of memory, so it was
> probably due to IPC and messaging.
> The general discussion on the Giraph mailing list strongly indicates that
> I did hit the current IPC scalability limits.
>
> Currently I am working on a non-Hadoop version of the algorithm which is
> not as scalable but which is fast for *one* user. ( less then a second per
> user, but single threaded).
> In addition, this new version allows me to better integrate with an
> existing ecosys

Re: Suggestions on problem sizes for giraph performance benchmarking

[Benjamin Heitmann]

Hello Stephen, 

sorry for the very late reply. 

On 28 Jun 2012, at 02:50, Fleischman, Stephen (ISS SCI - Plano TX) wrote:

> Hello Avery and all:
> I have a cluster of 10  two-processor/48 GB RAM servers, upon which we are 
> conducting Hadoop performance characterization tests.  I plan to use the 
> Giraph pagerank and simple shortest path example tests as part of this 
> exercise and would appreciate guidance on problem sizes for both tests.  I’m 
> looking at paring down an obfuscated Twitter dataset and it would save a lot 
> of time if someone has some knowledge on roughly how the time and memory 
> scales with number of nodes in a graph.
> 


I can provide some suggestions for the kind of algorithm and data which does 
currently surpass the scalability of giraph. 

While the limits to my knowledge of Giraph and Hadoop are probably also to 
blame for this, please see the recent discussions on this list, 
and on JIRA for other indications that the scalability of Giraph needs 
improvement: 
* post  by Yuanyuan Tian in the thread "wierd communication errors" on 
user@giraph.apache.org
* GIRAPH-234 about GC overhead 
https://issues.apache.org/jira/browse/GIRAPH-234?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel

If you want to stretch the limits of Giraph, then you need to try an algorithm 
which is conceptually different from PageRank, and you need a big data set. 
If you use an algorithm which has complex application logic (maybe even domain 
specific logic), which needs to be embedded in the algorithm, 
then the nodes need to have a lot of state. In addition, such algorithms 
probably send around a lot of messages, and each of the messages might have a 
payload
which is more complex then one floating point number. In addition, it helps to 
have a graph format, which requires strings on the edges and vertices. 
The strings are required for the domain specific business logic which the graph 
algorithm needs to follow. 

Finally, imagine a data set which has a big loading time, and where one run of 
the algorithm only provides results for one user.
The standard Hadoop paradigm is to throw away the graph after loading it. 
So if you have 100s or 1000s of users, then you need a way to execute the 
algorithm multiple times in parallel. 
Again this will add a lot of state, as each of the vertices will need to hold 
one state object for each user who has visited the vertex.

In my specific case, I had the following data and algorithm:
Data:  
* an RDF graph with 10 million vertices and 40 million edges 
I used my own import code to map the RDF graph to a undirected graph with a 
limit of one edge between any two nodes (so it was not a multi-graph)
* each vertex and each edge uses a string as an identity to represent a URI in 
the RDF graph (required for the business logic in the algorithm)

Algorithm: 
* spreading activation. 
You can think of it as depth first search guided by domain specific logic. 
A short introduction here: https://en.wikipedia.org/wiki/Spreading_activation
The wikipedia article only mentions using spreading activation on weighted 
graphs, however I used it on graphs which have additional types on the edges. 
The whole area of using the semantics of the edges to guide the algorithm is an 
active research topic, so thats why I can't point you to a good article on 
that. 
* parallel execution: 
I need to run the algorithm once for every user in the system, however loading 
the data set takes around 15 minutes alone. 
So each node has an array of states, one for each user for which the algorithm 
has visited a node. 
I experimented with user numbers between 30 and 1000, anything more did not 
work for concurrent execution of the algorithm. 

Infrastructure: 
* a single server with 24 Intel Xeon 2.4 GHz cpus and 96 GB of RAM
* Hadoop 1.0, pseudo-distributed setup
* between 10 and 20 Giraph workers


A few weeks ago I stopped work on my Giraph based implementation, as Giraph ran 
out of memory almost immediately after loading and initialising the data. 
I made sure that the Giraph workers do not run out of memory, so it was 
probably due to IPC and messaging. 
The general discussion on the Giraph mailing list strongly indicates that I did 
hit the current IPC scalability limits.

Currently I am working on a non-Hadoop version of the algorithm which is not as 
scalable but which is fast for *one* user. ( less then a second per user, but 
single threaded). 
In addition, this new version allows me to better integrate with an existing 
ecosystem of technologies (Semantic Web technologies) to which Hadoop and 
Giraph is currently completely disconnected. 
However, I will probably revisit Giraph at some time on the future. 


If you want to look at the code or the data or any other asset which I have, 
then I will gladly share that with you. 
I would really like Giraph to reach the maturity required for this kind of 
algorithm. 
However, I have the feeling that the current devel

Suggestions on problem sizes for giraph performance benchmarking

[Fleischman, Stephen (ISS SCI - Plano TX)]

Hello Avery and all:
I have a cluster of 10  two-processor/48 GB RAM servers, upon which we are 
conducting Hadoop performance characterization tests.  I plan to use the Giraph 
pagerank and simple shortest path example tests as part of this exercise and 
would appreciate guidance on problem sizes for both tests.  I'm looking at 
paring down an obfuscated Twitter dataset and it would save a lot of time if 
someone has some knowledge on roughly how the time and memory scales with 
number of nodes in a graph.

Best regards,
Steve Fleischman