Mailing-List: contact user-help@hbase.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@hbase.apache.org
Received-SPF: softfail (athena.apache.org: transitioning domain of
 abhishek1015@gmail.com does not designate 216.139.236.26 as permitted sender)
Date: Mon, 21 Jul 2014 14:19:50 -0700 (PDT)
From: abhishek1015 <abhishek1015@gmail.com>
To: user@hbase.apache.org
Message-ID: <1405977590187-4061748.post@n3.nabble.com>
Subject: insert latency
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

Hello everyone,

I have a two node hbase cluster with each node running a data-node and a
region-server. I am running YCSB benchmark on this cluster. I have these two
counter-intuitive following interesting observation which I could not
explain. Hoping someone could help understand these behavior.

Experiment 1: a) Write Only - 100% insert workload vs b) Write heavy - 90 %
insert and 10% read
Observation: While average insert latency for write heavy is around 15 us,
the average insert latency is about 70x to 120x more for Write only
operations

Experiment 2: c) Write only workload with default compaction scheme, vs d)
Write only  workload with major and minor compaction off
Observation: While less data is written for compaction off, the insert
latency is lesser for default setting.

Thanks for any help.

Abhishek 


--
View this message in context: http://apache-hbase.679495.n3.nabble.com/insert-latency-tp4061748.html
Sent from the HBase User mailing list archive at Nabble.com.