AWS Neptune's performance

734 views
Skip to first unread message

Ashwin Biyani

unread,
May 29, 2018, 12:56:45 AM5/29/18
to Gremlin-users
Hi All,

Does anyone have performance data of AWS Neptune? 

We have 1 million vertices and 3.5 million edges loaded. I am running a little complex query with a limit as 10, it takes around 0.7 seconds to iterate and get data from all the 10 output vertices.

Query:
Object[] vertexIds = {"1000000954618","1000001253078","1000001359440","1000001359441","1000001359442"};
GraphTraversal t =   g.V(vertexIds).as("source").outE("associated").has("lift",P.gt(5000)).order().
by("lift", Order.decr).as("lift").inV().as("target").select("source","target","lift"). by("product_code").
by("product_code").
by("lift").limit(10);
 

Tried the same query with 100 concurrent requests on JMeter and it takes 9 seconds. Does anyone else have their findings on Neptune. Are we missing something here?
How does Neo4J's performance on similar infrastructure?

Thanks,
Ashwin


Stephen Mallette

unread,
May 29, 2018, 7:48:08 AM5/29/18
to Gremlin-users
Tried the same query with 100 concurrent requests on JMeter and it takes 9 seconds.

what takes 9 seconds? 100 requests total all sent at the same time? or each request from jmeter takes 9 seconds? something else?

How does Neo4J's performance on similar infrastructure?

Hard to answer that. I mean, 3.5 million edges is a nice size for Neo4j because the whole graph can fit in memory really easily. Once you warm up neo4j caches and get that graph in-memory, it's going to have no problems being "fast". I'm not an expert on Neptune, but I sense that like other graphs of its kind, it would do better when your graph could no longer fit in memory anymore. 



--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/55a8eb10-5ea1-49bd-9617-b24a96eee5a3%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Ashwin Biyani

unread,
May 29, 2018, 7:51:26 AM5/29/18
to gremli...@googlegroups.com
100 request in total takes 9 seconds get the results and iterate and display. Is not it on higher side. 

On Tue, May 29, 2018, 17:18 Stephen Mallette <spmal...@gmail.com> wrote:
Tried the same query with 100 concurrent requests on JMeter and it takes 9 seconds.

what takes 9 seconds? 100 requests total all sent at the same time? or each request from jmeter takes 9 seconds? something else?

How does Neo4J's performance on similar infrastructure?

Hard to answer that. I mean, 3.5 million edges is a nice size for Neo4j because the whole graph can fit in memory really easily. Once you warm up neo4j caches and get that graph in-memory, it's going to have no problems being "fast". I'm not an expert on Neptune, but I sense that like other graphs of its kind, it would do better when your graph could no longer fit in memory anymore. 


On Tue, May 29, 2018 at 12:56 AM, 'Ashwin Biyani' via Gremlin-users <gremli...@googlegroups.com> wrote:
Hi All,

Does anyone have performance data of AWS Neptune? 

We have 1 million vertices and 3.5 million edges loaded. I am running a little complex query with a limit as 10, it takes around 0.7 seconds to iterate and get data from all the 10 output vertices.

Query:
Object[] vertexIds = {"1000000954618","1000001253078","1000001359440","1000001359441","1000001359442"};
GraphTraversal t =   g.V(vertexIds).as("source").outE("associated").has("lift",P.gt(5000)).order().
by("lift", Order.decr).as("lift").inV().as("target").select("source","target","lift"). by("product_code").
by("product_code").
by("lift").limit(10);
 

Tried the same query with 100 concurrent requests on JMeter and it takes 9 seconds. Does anyone else have their findings on Neptune. Are we missing something here?
How does Neo4J's performance on similar infrastructure?

Thanks,
Ashwin


--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CAA-H439WF76qs4z0xx-ggmmKGbh3YEZtGMOnWSGP05uJ3mZRqw%40mail.gmail.com.

pieter gmail

unread,
May 29, 2018, 11:46:54 AM5/29/18
to gremli...@googlegroups.com
Hi,

If you have some code that creates a sample graph then I'll be happy to try out your query on Sqlg.

Cheers
Pieter

Bryan B. Thompson

unread,
May 29, 2018, 3:03:13 PM5/29/18
to Gremlin-users
Ashwin,

Can I get a copy of the dataset?

Thanks,
Bryan

Stephen Mallette

unread,
May 29, 2018, 3:08:57 PM5/29/18
to Gremlin-users
Hi Bryan, I think i recognize your name - you work on Neptune, don't you? 

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/d75466ac-3fae-4306-9da1-98043ad5eceb%40googlegroups.com.

Ashwin Biyani

unread,
May 29, 2018, 3:15:26 PM5/29/18
to gremli...@googlegroups.com
Hi,
Is this Bryan from AWS team(got from linkedln) . If so, I have already shared account and endpoint details(data is already loaded here) in an email to Divij vaidya, Brad bebe of Neptune team. 

Bryan B. Thompson

unread,
May 29, 2018, 3:16:47 PM5/29/18
to gremli...@googlegroups.com
Yes.  I will follow up with Brad & Divij.

Thanks,
Bryan

To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-users+unsubscribe@googlegroups.com.

--
You received this message because you are subscribed to a topic in the Google Groups "Gremlin-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gremlin-users/t_4T5fw-UZ8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to gremlin-users+unsubscribe@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/gremlin-users/CABjaGU%3DTbyg5%3D%3DZuH1Nt0Q5qW0ZcxEK0Q-%2BQudBMbEc53DtqkA%40mail.gmail.com.

Stephen Mallette

unread,
May 29, 2018, 3:23:16 PM5/29/18
to Gremlin-users
Bryan - nice to know you're following along here. I think we see a fair share of Neptune questions on this list, so good to see some devs are following along. Thanks.

On Tue, May 29, 2018 at 3:16 PM, Bryan B. Thompson <thomp...@gmail.com> wrote:
Yes.  I will follow up with Brad & Divij.

Thanks,
Bryan

Ashwin Biyani

unread,
May 29, 2018, 3:33:10 PM5/29/18
to gremli...@googlegroups.com
Thanks. Responded to your email with further details. 

--
You received this message because you are subscribed to a topic in the Google Groups "Gremlin-users" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/gremlin-users/t_4T5fw-UZ8/unsubscribe.
To unsubscribe from this group and all its topics, send an email to gremlin-user...@googlegroups.com.

--
You received this message because you are subscribed to the Google Groups "Gremlin-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to gremlin-user...@googlegroups.com.

pieter gmail

unread,
May 29, 2018, 3:45:16 PM5/29/18
to gremli...@googlegroups.com
Hi,

I am able to attach "2G_Dover_Rd_SGC" boards successfully.

However to "2G_Swiss_Club_SGC" it is failing. I can find/create all the board equipment but attaching it to the parent site equipment fails.

The url is

https://aina033f99bd-a4a5dfb44.hana.ondemand.com/ain/services/api/v1/equipment%286EAE57511C714B42917308FD57EC9363%29/components?type=EQU

and the PUT payload is,

{"children":[{"id":"0074685DDF0440BA8AE211DF79D739FE","order":"1","objectType":"EQU"},{"id":"37FF2401EC71421886B0FFC0C41343ED","order":"2","objectType":"EQU"},{"id":"887983F9252B4C7B946353ADEB8F8961","order":"3","objectType":"EQU"},{"id":"B461D7C3496348C58BCF67B3AE5FFD18","order":"4","objectType":"EQU"},{"id":"56D41AE8CABB46048DB7373C68417B90","order":"5","objectType":"EQU"},{"id":"D74EC76C66B146EC843E275AAB78736B","order":"6","objectType":"EQU"},{"id":"E5FA55F0F6E848458095B2231E4B6CE1","order":"7","objectType":"EQU"}]}

and the error is,

Error attaching equipment [0074685DDF0440BA8AE211DF79D739FE, 37FF2401EC71421886B0FFC0C41343ED, 887983F9252B4C7B946353ADEB8F8961, B461D7C3496348C58BCF67B3AE5FFD18, 56D41AE8CABB46048DB7373C68417B90, D74EC76C66B146EC843E275AAB78736B, E5FA55F0F6E848458095B2231E4B6CE1] to virtual site equipment 6EAE57511C714B42917308FD57EC9363. Error code 404. Error body {"errorMessages":[{"severity":"ERROR","errorMessage":"The system could not find the equipment.","messageCode":"EQU401"}],"statusCode":404}

Thanks
Pieter
Reply all
Reply to author
Forward
0 new messages