> Actually, something I just came across and hadn't anticipated.

> The repository call is actually pretty quick; on the order of several
> seconds (like no more than 5-10 tops).

> While it could be better (given that it's only 550 nodes being returned),
> it's sufficiently fast for now.

> What's slowing things down is this: My conversion of findAll()'s return
> type (i.e. a ClosableIterable) into a Collection (via
> IteratorUtil.asCollection(...) is what's taking up the majority of the time.

> So, this is either something I've forgotten about since my hardcore Java
> days, or I'm going to have to look into why the conversion is so horribly
> inefficient (I do get the differences between the Iterable concept and the
> Collection concept).  I'm using a "Collection" since I'm assigning it to a
> "select" box in my JSP via Spring MVC.

> Thoughts are welcome (I've probably overlooked something obvious).

> So, long story short: The timing code I need can likely wait until I have
> things straightened out a bit more.  I would love to see my query come down
> from ~5s to a lot less.

> BTW, I found that the findAll() method was just as quick (if not quicker)
> than my use of the same Cypher query and the use of @MapResult, etc.

> Thanks again guys!  Very excited going through all this!

> On Friday, September 28, 2012 3:55:29 PM UTC-4, BtySgtMajor wrote:

>> Hi Michael,

>>   What you're proposing in terms of adding time logging for repos/queries
>> is exactly what I'm looking for.  I don't have any eager loading in place
>> at this point in time anyway (it's the last thing I'd want to add).

>>   I'll give the logging proxy bit a look as I have a chance.  As for the
>> filter option, that might take me a bit longer, and will give that a shot
>> if it comes to that.

>>   With respect to my code, there's really not all that much.

>>   Essentially, from within one of my Spring MVC controllers, I have a
>> line like this:  ("Game" is part of my domain model as I'm doing something
>> similar to IMDB.)

>> Collection<Game> allGames =
>> IteratorUtil.asCollection(gameRepository.findAll());

>>   And, the "gameRepository" is declared like this:

>> public interface GameRepository extends GraphRepository<Game>,
>> RelationshipOperationsRepository<Game> {
>> Game findByGameId(String id);
>> }

>>   That's about it.  Just keeping things simple.

>>   The "Game" model does have a few relationships; however, all
>> relationships/collections are left as lazily loaded so as to avoid a big
>> mess.

>>   Currently, running the code above takes, on average, about 35s to
>> complete the round trip.

>>   If I run the same query via Cypher on the Neo4j console, it takes a
>> whopping 42ms.

>>   Also, if I just fetch a single game via the gameId via SDN, it takes
>> like 600ms.

>>   So, like I said, early days, and I'm trying to really dig in and find
>> out exactly where bottlenecks are occurring so I can rip it apart and
>> really get familiar with it.  Like I said before, I have my hypotheses, but
>> I need ways to test them out. :)

>>   Let me know if you want/need additional info/code and, if so, what it
>> is.

>>   Thanks again, Michael!

>> Cheers,
>> Duncan

>> On Friday, September 28, 2012 2:01:30 PM UTC-4, Michael Hunger wrote:

>>> Duncan,

>>> that is a good point actually.

>>> I can add time logging to the repositories / queries in SDN. Due to the
>>> lazy nature that will be only the query parsing and initial execution
>>> though (might be the full query time though depending on if you have a
>>> streaming or non-streaming server interaction).

>>> Can you share some of the code you're talking about?

>>> An more or less easy way to do it for you might be putting a
>>> logging-proxy in front of the neo4j server

>>> Alternatively if you're more adventurous adding a filter (via an
>>> unmanaged extension) to the neo4j server that logs the request times.
>>> (see:
>>> https://github.com/neo4j/authentication-extension/blob/1.8/src/main/j...,
>>> also make sure to have the META-INF/services/.... with your initializer
>>> class name) there.

>>> HTH

>>> Michael

>>> Am 28.09.2012 um 08:26 schrieb BtySgtMajor:

>>> Hi all,

>>>   So I'm getting into some good stuff here as I dig into Neo4j deeper.
>>>  At present, I'm trying to determine where bottlenecks lie as I develop
>>> some SDN code around a Neo4j server.  Obviously, I can time the Java code
>>> that makes the call out to the server (which is remote--on the same
>>> network, but sitting on a VM on another box), and I can see the time it
>>> takes my queries through the Neo4j console.

>>>   My question is this: Is there a way in the server/logging config in
>>> Neo4j to document/log each query's runtime?  This would help me greatly in
>>> determining where my bottlenecks are.  (I'd imagine it's in the part where
>>> my SDN repository is calling out to the remote Neo4j server, but I'd like
>>> to confirm and then start tweaking things).

>>>   Thanks in advance!  Exciting times!

>>> Cheers,
>>> Duncan

>>> --