Hi Michael
Yes, I'm currently using 2.0. 
But since the app logic allows me to express the communication with the DB 
in terms of retrieving nodes by their ids, I don't have to ineluctably use 
indexing (as well as labels at all), so my current solution is to disable 
the indexing at all. Anyway, it all was just playing around and running 
some tests. I'm going to dedicate the next week (or the week after 
holidays) to testing the cluster setup with 3 nodes running standalone 
servers (got an enterprise edition finally) + some stress-tests like 
switching off 1 or 2 nodes, switching them off AND "losing" the data 
intentionally and so on. I'm planning to increase the amount of nodes to, 
perhaps, 1 billion and see what it shows.
Let me know if you are interested in updates so I could keep you posted.
And about the second part of you post: that's how 350M nodes look on the 
SSD now (indices disabled, single-node setup, two string properties per 
node, on relationship per node, tree structure):
namei@host:~|=>  du -s /ssd/neotree/* | sort -g -r
26443856        /ssd/neotree/neostore.propertystore.db
10645264        /ssd/neotree/neostore.relationshipstore.db
4514808 /ssd/neotree/neostore.nodestore.db
25604   /ssd/neotree/nioneo_logical.log.v1
25604   /ssd/neotree/nioneo_logical.log.v0
3932    /ssd/neotree/nioneo_logical.log.v2
392     /ssd/neotree/messages.log
36      /ssd/neotree/index
32      /ssd/neotree/tm_tx_log.2
20      /ssd/neotree/schema

Regards
Roman

On Thursday, December 26, 2013 10:59:30 PM UTC+2, Michael Hunger wrote:
>
> Hi Roman,
>
> I assume you use 2.0 ?
>
> There are some issues with building the label scan store with the 2.0 
> batch inserter.
>
> Something you can do is to use the 1.9 batch-inserter to add your nodes 
> then upgrade the store, and add labels after the fact using the 
> transactional db (if need be).
>
> If you don't use labels for now you can also stay with the 1.9 version so 
> long.
>
> What about relationships? How many do you have and do they fit on the SSD 
> as well?
>
> HTH
>
> Michael
>
> Am 26.12.2013 um 12:48 schrieb Roman Tkalenko <[email protected]<javascript:>
> >:
>
> Michael,
> well, that's a pity. Thank you for an answer anyway.
>
> I don't want to start a new thread, so I'll ask another question right 
> here:
> I'm using BatchInserter to insert ~500 000 000 nodes into the db and they 
> fit onto a 60 GB SSD nicely, but after the insertion finishes, indexing 
> starts and the whole process fails because of `no place left on the 
> device`. I didn't create index (i didn't assign any labels to nodes) and 
> didn't call anything like `createDeferredSchemaIndex()` or so on. Why is 
> the indexing process being launched anyway? How do I completely disable it?
> Thank you,
> Roman.
>
> On Thursday, December 26, 2013 8:11:05 AM UTC+2, Michael Hunger wrote:
>>
>> Roman,
>>
>> unfortunately no to 1) and 2) but this is an interesting suggestion.
>>
>> Currently the two modes only differ in allowing writes or not, not much 
>> in terms of performance (i.e. removing locks or synchronization on the r/o 
>> connection)
>>
>> Michael
>>
>> Am 25.12.2013 um 16:16 schrieb Roman Tkalenko <[email protected]>:
>>
>> Hi,
>> I'd like to hear your advice concerning the following:
>>
>>    1. is it even possible - open two simultaneous connections to 
>>    (already existing) database, one of which will be opened as 
>>    "read-only", and the other one - with full access?
>>    2. is it reasonable and will my app gain significant performance 
>>    under the condition that I have more read-only operations than r-w?
>>    3. if answers to 1) and 2) are "yes"? than what is the most 
>>    appropriate way to do it?
>>
>> Thank you,
>> Roman.
>>
>> -- 
>> You received this message because you are subscribed to the Google Groups 
>> "Neo4j" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to [email protected].
>> For more options, visit https://groups.google.com/groups/opt_out.
>>
>>
>>
> -- 
> You received this message because you are subscribed to the Google Groups 
> "Neo4j" group.
> To unsubscribe from this group and stop receiving emails from it, send an 
> email to [email protected] <javascript:>.
> For more options, visit https://groups.google.com/groups/opt_out.
>
>
>

-- 
You received this message because you are subscribed to the Google Groups 
"Neo4j" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/groups/opt_out.

Reply via email to