[
https://issues.apache.org/jira/browse/PHOENIX-971?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Nick Dimiduk updated PHOENIX-971:
---------------------------------
Attachment: image-2.png
I've made some good progress with Calcite's Avatica and have some preliminary
results I'm able to share. All the work is on my pair of branches, for
[calcite|https://github.com/ndimiduk/incubator-calcite/tree/calcite-618] and
[phoenix|https://github.com/ndimiduk/phoenix/tree/WIP-971-Avatica]. The result
is a query server that runs json over http, and thin client implemented as
JDBC. I have a script for launching the query server, and a new sqlline-thin.py
script that uses this thin client.
All this i've been testing using JMeter and a TCP-DS dataset that small enough
to sit in memory -- intended to test the server infra alone. I've attached some
preliminary performance numbers from local machine testing in image-2.png. The
Avatica server has pretty decent performance, within 15% of the "thick" JDBC
client for the single process test (and it's actually out-performing the thick
client with multiple client processes... which doesn't make sense to me).
Site note: I'm having a bit of trouble with extending the Phoenix end2end suite
to add coverage for this query server. Please give me some hits via the thread
I sent to dev@.
Feedback welcome!
> Query server
> ------------
>
> Key: PHOENIX-971
> URL: https://issues.apache.org/jira/browse/PHOENIX-971
> Project: Phoenix
> Issue Type: New Feature
> Reporter: Andrew Purtell
> Assignee: Nick Dimiduk
> Fix For: 5.0.0
>
> Attachments: image-2.png
>
>
> Host the JDBC driver in a query server process that can be deployed as a
> middle tier between lighter weight clients and Phoenix+HBase. This would
> serve a similar optional role in Phoenix deployments as the
> [HiveServer2|https://cwiki.apache.org/confluence/display/Hive/Setting+Up+HiveServer2]
> does in Hive deploys.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
