Can you please comment in the pagecache setting and put it (it close to the sum of neostore.*.db) files but keep the total memory and heap into account
Von meinem iPhone gesendet > Am 13.03.2017 um 08:54 schrieb unrealadmin23 via Neo4j > <[email protected]>: > > I have the following settings in neo4j.conf: > > This is about 1/8th of the machines memory: > > dbms.memory.heap.initial_size=120000m > dbms.memory.heap.max_size=120000m > > > The page cache is commented out: > #dbms.memory.pagecache.size=10g > > Wayne > > >> On Monday, 13 March 2017 02:35:00 UTC, Michael Hunger wrote: >> What does your configuration look like in neo4j.conf for heap and page-cache? >> >> neo4j> call apoc.warmup.run() ; >> pageSize nodesPerPage nodesTotal nodePages nodesTime >> relsPerPage relsTotal >> 8192 546 123396013 226001 2 >> 240 2035429925 >> relPages relsTime totalTime >> 8480959 73 76 >> >> so it loaded 226001 pages of nodes totalling 123m = 1.7 GB >> and 8480959 pages of relationships totalling 2bn = 65 GB >> >> in 76 seconds >> >> So depending on your configuration you either have now 67GB of page-cache >> filled or it spilled over. >> >> All operations that now require only node or relationship-record access >> should access them in-memory. >> >> Operations that require properties will still need to load them into the >> page-cache and depending on the size setup displace existing entries. >> >> >> >>> On Sun, Mar 12, 2017 at 4:45 PM, unrealadmin23 via Neo4j >>> <[email protected]> wrote: >>> I installed apoc - I didn't know this existed - looks very comprehensive; >>> thanks: >>> >>> neo4j> call apoc.warmup.run() ; >>> pageSize, nodesPerPage, nodesTotal, nodePages, nodesTime, relsPerPage, >>> relsTotal, relPages, relsTime, totalTime >>> 8192, 546, 123396013, 226001, 2, 240, 2035429925, 8480959, 73, 76 >>> >>> Not much in the way of a noticeable speedup though. >>> >>> I don't know anything about Java, and cannot find the jvm command. Can >>> you be more specific about what I need to run or configure from the Linux >>> command line ? >>> >>> Wayne. >>> >>> >>>> On Saturday, 11 March 2017 10:14:49 UTC, Michael Hunger wrote: >>>> There is not a kernel buffer, but own own page cache that you have to >>>> configure >>>> >>>> call apoc.warmup.run() >>>> >>>> Which warms up node and rel blocks >>>> >>>> Other speedup comes from jvm JIT warmup >>>> >>>> >>>> Von meinem iPhone gesendet >>>> >>>>> Am 11.03.2017 um 08:28 schrieb unrealadmin23 via Neo4j >>>>> <[email protected]>: >>>>> >>>>> Hi, >>>>> >>>>> Assuming that I have enough memory to loaded the entire DB into it, are >>>>> there any options (optimisations) to allow for this or is it just a case >>>>> of the disk blocks being able to reside in the kernels buffer cache ? >>>>> >>>>> Currently, the same query gets faster the more it is run - to a point. >>>>> I can live with an elongated startup time. >>>>> >>>>> Thanks, Wayne >>>>> -- >>>>> 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. >>> >>> -- >>> 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. >> > > -- > 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. -- 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.
