Hacker News Reader

A pity that the paper doesn't include Hive LLAP in the comparison. With LLAP enabled, Hive becomes more like Impala and Drill, with persistent daemon processes providing much shorter response times to queries.

I'm curious how Hive LLAP does, compared to Impala and Drill.

View on HN

I found that a curious omission too -- it's hardly possible to have a meaningful study of SQL-on-Hadoop interactive analytics in 2017 (conservatively assuming there is a 1-year lag between the study and the paper) without including Hive LLAP and Presto.

But having been part of the academic machine, I kind of understand. These things happen.

EDIT: I notice that a short paper was submitted to IEEE in 2017, so the study was probably done in 2016, which may explain the omission.

There're better comparisons put out by vendors who are building custom solutions on top of the open source Hadoop SQL engines, and from Hortonworks/Cloudera as well.

Hortonworks has been pushing their latest Hive LLAP benchmarks pretty hard. They're doing a kind of a Hive3 "roadshow" at the moment as well.

Reply on news.ycombinator.com