Update: after creating an index on property of type integer, fetching a node by id seems it take same time as by name (fulltext index): ~6K ms create index on:topic(id)
(note here id is not the interna lnode identifier of neo) MATCH (n) WHERE n.id =9996533 Return n; ~6K ms while with internal id MATCH (n) WHERE id(n) =467383 Return n; 234 ms On Fri, Oct 3, 2014 at 3:10 PM, XDiscovery Team <[email protected]> wrote: > Oh I see, I did > > create index on:topic(name) > > and read your useful post > > > Again questions about the confusion: > > in your post you wrote i should not mix the indexes, and that if i need to > use fulltext, then legacy indexes are the one to use. > > 1. Could you please brief me on the proper index to set (index legacy Vs > schema) for a product like: > > - ability to fetch a node and its relationships / paths, based on its > names. > I can set names on node properties, they could be a single one (e.g. > 'Italy') or multiple once (e.g. 'Italy', 'Italie', etc.) > > - nodes are all of the same types (one label), at least for large sets > (e.g. millions) > > 2. Since I have many languages to name a node, what would you suggest, > keeping in mind i have to perform fulltext search: > - a list of nodes' properties by language > - an array of property names [IT; DE, EN, FR...] > - other nodes, as many as the names associted to that name are, for each > node indexsing its namel (in this case I would have a graph of 4M nodes* > *m* languages to index) > > 3. How coudl improve responsiveness, even when using schema ? > > As example, here some metrics: > > - legacy index and not schema > MATCH n-[r]-m where n.name = 'tittle' return m limit by 6 > ~126K ms > > MATCH [..] query to identify all shortest paths between two nodes: > ~600K ms > > > -legacy index with schema [create index on:topic(name) ] : > MATCH n-[r]-m where n.name = 'tittle' return m limit by 6 > ~6K ms > > MATCH [..] query to identify all shortest paths between two nodes: > ~18K ms > > How to reduce time for fetching a node and its realtionships, paths to ms > and not Kms ? (production level) > > Here using a laptop 8GB RAM, 6GB dedicated to JVM > > thank you very much Micheal, it was not simple to find out this > information; I read fulltext performance is on roadmap, but would need to > understand if my data structure is ok and i am on the right path: MATCH > me-[deployment]-[MVP]-neo4J :D > > > > > > > > > On Fri, Oct 3, 2014 at 1:27 PM, Michael Hunger < > [email protected]> wrote: > >> create index on :topic(name); > > > > > > -- > *Luigi Assom* > Founder & CEO @ XDiscovery - Crazy on Human Knowledge > www.xdiscovery.com | http://learn.xdiscovery.com > T +39 349 3033334 > E [email protected] > Skype oggigigi > -- Luigi Assom Skype contact: oggigigi -- 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/d/optout.
