> > we experimented with issuing the same query repeatedly, and we observed a > slow down not only on the first query
I am not sure what the reasons are, perhaps you can enable TRACE log to find what leads to slow, I guess that some meta information is reloaded under highly write workload. ---------------------------------------- Jaanai Zhang Best regards! William Shen <wills...@marinsoftware.com> 于2019年2月1日周五 上午2:09写道: > Thanks Jaanai. Do you know if that is expected only on the first query > against a table? For us, we experimented with issuing the same query > repeatedly, and we observed a slow down not only on the first query. Does > it make sense to preemptively load table metadata on start up to warm up > the system to reduce latency during the actual query time (if it is > possible to do so)? > > On Wed, Jan 30, 2019 at 10:54 PM Jaanai Zhang <cloud.pos...@gmail.com> > wrote: > >> It is expected when firstly query tables after establishing the >> connection. Something likes loads some meta information into local cache >> that need take some time, mainly including two aspects: 1. access >> SYSTEM.CATALOG table to get schema information of the table 2. access the >> meta table of HBase to get regions information of the table >> >> ---------------------------------------- >> Jaanai Zhang >> Best regards! >> >> >> >> William Shen <wills...@marinsoftware.com> 于2019年1月31日周四 下午1:37写道: >> >>> Hi there, >>> >>> I have a component that makes Phoenix queries via the Phoenix JDBC >>> Connection. I noticed that consistently, the Phoenix Client takes longer to >>> execute a PreparedStatement and it takes longer to read through the >>> ResultSet for a period of time (~15m) after a restart of the component. It >>> seems like there is a warmup period for the JDBC connection. Is this to be >>> expected? >>> >>> Thanks! >>> >>