The overhead of just opening a core is insignificant relatively to using it so, unless you are worried about hitting the max number of open files limit, it seems unimportant.
Otis Solr & ElasticSearch Support http://sematext.com/ On Apr 23, 2013 7:46 AM, "Jérôme Étévé" <jerome.et...@gmail.com> wrote: > Thanks! > > Yeah I know about the caching/commit things > > My question is more about the impact of a "Pure" creation of a Solr core, > indepently of its "usage" memory requirements (like caches and stuff). > > From the experiments I did using JMX, it's not measurable, but I might be > wrong. > > > > On 23 April 2013 12:25, Guido Medina <guido.med...@temetra.com> wrote: > > > I'm not an expert, but at some extent I think it will come down to few > > factors: > > > > * How much data is been cached per core. > > * If memory is an issue and still you want performance, I/O with low > > cache could be an issue (SSDs?) > > * Soft commits which implies open searchers per soft commit (and per > > core) will depend on caches. > > > > I do believe at the end it will be a direct result of your caching and > > I/O. If all you care is performance, caching (memory) could be replaced > > with faster I/O though soft commits will be fragile to memory due to its > > nature (depends on caching/memory and low I/O usage) > > > > Hope I made sense, I probably tried too many points of view in a single > > idea. > > > > Guido. > > > > > > On 23/04/13 11:50, Jérôme Étévé wrote: > > > >> Hi all, > >> > >> We've got quite a lot of (mostly small) solr cores in our Solr instance. > >> They all share the same solrconfig.xml and schema.xml (only the data > >> differs). > >> > >> I'm wondering how far can I go in terms of number of cores. CPU is not > an > >> issue, but memory could be. > >> > >> An idea/guideline about the impact of a new Solr Core in a Solr > instance? > >> > >> Thanks! > >> > >> Jerome. > >> > >> > > > > > -- > Jerome Eteve > +44(0)7738864546 > http://www.eteve.net/ >
