phoenix-user mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Amit Shah <amits...@gmail.com>
Subject Understanding Phoenix Query Plans
Date Mon, 11 Apr 2016 10:09:47 GMT
Hi,

I am using hbase version 1.0 and phoenix version 4.6. For different queries
that we are benchmarking, I am trying to understand the query plan

1. If we execute a where clause with group by query on primary key columns
of the table, the plan looks like
+------------------------------------------+
|                   PLAN                   |
+------------------------------------------+
| CLIENT 1-CHUNK PARALLEL 1-WAY POINT LOOKUP ON 1 KEY OVER TRANSACTIONS |
+------------------------------------------+

What does "1-way point lookup on 1 key" mean?

2. When executing a where clause query with group by on non primary key
columns, the plan looks like

+------------------------------------------+
|                   PLAN                   |
+------------------------------------------+
| CLIENT 48-CHUNK PARALLEL 48-WAY FULL SCAN OVER TRANSACTIONS |
|     SERVER FILTER BY CH_TYPE = 'CHANNEL_TYPE_32' |
|     SERVER AGGREGATE INTO DISTINCT ROWS BY [T_COUNTRY] |
| CLIENT MERGE SORT                        |
+------------------------------------------+

A full table scan is being performed.
Does "CLIENT" and "SERVER" refer to the operations performed on the client
(which is sqlline) and the server (hbase region server) ?
Does "48-WAY" indicate that the query is split across region servers and
executed in multiple (in this case 48) threads? If so, what config controls
this concurrency?


Thanks,
Amit.

Mime
View raw message