Wow.  That's surprising data.  With Neo4J embedded, we can usually get about 
100X of that performance (including Lucene indexing of the nodes), so there 
clearly seem to be some big impacts from using the REST interface versus Neo4J 
embedded.

-----Original Message-----
From: user-boun...@lists.neo4j.org [mailto:user-boun...@lists.neo4j.org] On 
Behalf Of jadell
Sent: Sunday, August 28, 2011 3:55 PM
To: user@lists.neo4j.org
Subject: [Neo4j] More batch vs. single operation benchmarks

Here are the latest benchmarks of batch vs. individual entity creation using
the Neo4jPHP library.  Most of the processing time is spent on the server,
so I believe that these numbers are probably not specific to Neo4jPHP.  I'm
not implying that there is anything wrong or to be fixed; I just thought the
results might be of interest to others.  I'd love to see results from others
using a REST client in any language.

3 scenarios were run 10 times each for different batch sizes.  The scenarios
and averages of the 10 runs of each batch size are below.  First column is
the batch size, second column is the average time in seconds to create that
many entities in a batch, third column is the average time in seconds to
create that many entities with individual calls.

Benchmark script can be found here: http://gist.github.com/1177100


Results:

Scenario 1: Create nodes
size    batch    single
10      0        0
100     0        0.8
250     0.1      1.5
500     0.9      1.7
1000    1.5      3.8
2500    6.7      9.6
5000    23.5     13.2


Scenario 2: Create relationships
size    batch    single
10      0        0
100     0.2      0.2
250     0.3      0.5
500     0.6      1.1
1000    1.3      3
2500    7.1      9.7
5000    25.3     22.2


Scenario 3: Create 2 nodes and a relationship between them
size    batch    single
10      0        0.1
100     0.4      1.3
250     0.8      3
500     2.9      5.2
1000    9.2      11.9
2500    54.3     29.9
5000    710.3    59.6


Note: in the 5000 run of the last scenario, 15000 operations are sent in a
single HTTP request.  10 runs of the 5000 batch size (150000 operations
total operations in 10 batches) took almost 2 hours to complete.


-- Josh Adell


--
View this message in context: 
http://neo4j-community-discussions.438527.n3.nabble.com/More-batch-vs-single-operation-benchmarks-tp3291092p3291092.html
Sent from the Neo4j Community Discussions mailing list archive at Nabble.com.
_______________________________________________
Neo4j mailing list
User@lists.neo4j.org
https://lists.neo4j.org/mailman/listinfo/user
_______________________________________________
Neo4j mailing list
User@lists.neo4j.org
https://lists.neo4j.org/mailman/listinfo/user

Reply via email to