re: Dumb querying, Impala vs Hive
Slightly more impressively it just took 20seconds to tell me one of our tables has got 321842374 in it (that's ~321m). Not bad.Impala took slightly more than a second.
View Articlere: Dumb querying, Impala vs Hive
Yup. We got 0.13 just this morning in fact. Running count(*) on a ~9000 row table, 35s. Much better. Still not in Impala territory though.
View Articlere: Dumb querying, Impala vs Hive
I just ran a count(*) on a 220k row table and it ran in ~18 seconds.http://i.imgur.com/PdnZeLv.pngLooks like Stinger delivered in v0.13!
View Articlere: Dumb querying, Impala vs Hive
Yep, spot on Joe. Even the exec of the MR part took 23s. This shouldn't be a surprise of course, just interesting to see the difference exhibited so obviously.
View Articlere: Dumb querying, Impala vs Hive
I suspect/guess that's because even though it's empty a job is still launched along with a map task and the time you see is due to the task launching and completing. Impala doesn't convert your query...
View Article