From user-return-29274-apmail-cassandra-user-archive=cassandra.apache.org@cassandra.apache.org Tue Oct 2 13:18:25 2012 Return-Path: X-Original-To: apmail-cassandra-user-archive@www.apache.org Delivered-To: apmail-cassandra-user-archive@www.apache.org Received: from mail.apache.org (hermes.apache.org [140.211.11.3]) by minotaur.apache.org (Postfix) with SMTP id 16438DCF3 for ; Tue, 2 Oct 2012 13:18:25 +0000 (UTC) Received: (qmail 5656 invoked by uid 500); 2 Oct 2012 13:18:22 -0000 Delivered-To: apmail-cassandra-user-archive@cassandra.apache.org Received: (qmail 5583 invoked by uid 500); 2 Oct 2012 13:18:22 -0000 Mailing-List: contact user-help@cassandra.apache.org; run by ezmlm Precedence: bulk List-Help: List-Unsubscribe: List-Post: List-Id: Reply-To: user@cassandra.apache.org Delivered-To: mailing list user@cassandra.apache.org Received: (qmail 4646 invoked by uid 99); 2 Oct 2012 13:18:21 -0000 Received: from athena.apache.org (HELO athena.apache.org) (140.211.11.136) by apache.org (qpsmtpd/0.29) with ESMTP; Tue, 02 Oct 2012 13:18:21 +0000 X-ASF-Spam-Status: No, hits=2.2 required=5.0 tests=HTML_MESSAGE,RCVD_IN_DNSWL_NONE,SPF_PASS X-Spam-Check-By: apache.org Received-SPF: pass (athena.apache.org: domain of roshni_rajagopal@hotmail.com designates 65.55.34.142 as permitted sender) Received: from [65.55.34.142] (HELO col0-omc3-s4.col0.hotmail.com) (65.55.34.142) by apache.org (qpsmtpd/0.29) with ESMTP; Tue, 02 Oct 2012 13:18:15 +0000 Received: from COL121-W22 ([65.55.34.137]) by col0-omc3-s4.col0.hotmail.com with Microsoft SMTPSVC(6.0.3790.4675); Tue, 2 Oct 2012 06:17:54 -0700 Message-ID: Content-Type: multipart/alternative; boundary="_30e76393-e0c6-4a74-b8dc-34347e446a2d_" X-Originating-IP: [122.179.96.213] From: Roshni Rajagopal To: Subject: RE: Read latency issue Date: Tue, 2 Oct 2012 18:47:53 +0530 Importance: Normal In-Reply-To: References: <17C39FE466076C46B6E83F129C7B19CE16B6D515@HKXPRD0310MB352.apcprd03.prod.outlook.com>, MIME-Version: 1.0 X-OriginalArrivalTime: 02 Oct 2012 13:17:54.0447 (UTC) FILETIME=[54F9A1F0:01CDA0A0] X-Virus-Checked: Checked by ClamAV on apache.org --_30e76393-e0c6-4a74-b8dc-34347e446a2d_ Content-Type: text/plain; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable Arindam=2C Did you also try the cassandra stress tool & compare results? I havent done a performance test as yet=2C the only ones published on the i= nternet are of YCSB on an older version of apache cassandra=2C and it doesn= 't seem to be actively supported or updatedhttp://www.brianfrankcooper.net/= pubs/ycsb-v4.pdf.=20 The numbers you have sound very low=2C for a read of a row by key which sho= uld have been the fastest. I hope someone can help investigate or share nu= mbers from their tests. =20 Regards=2CRoshni=20 > From: Dean.Hiller@nrel.gov > To: user@cassandra.apache.org > Date: Tue=2C 2 Oct 2012 06:41:09 -0600 > Subject: Re: Read latency issue >=20 > Interesting results. With PlayOrm=2C we did a 6 node test of reading 100= rows from 1=2C000=2C000 using PlayOrm Scalable SQL. It only took 60ms. M= aybe we have better hardware though??? We are using 7200 RPM drives so not= hing fancy on the disk side of things. More nodes puts at a higher through= put though as reading from more disks will be faster. Anyways=2C you may w= ant to play with more nodes and re-run. If you run a test with PlayOrm=2C = I would love to know the results there as well. >=20 > Later=2C > Dean >=20 > From: Arindam Barua > > Reply-To: "user@cassandra.apache.org" <= user@cassandra.apache.org> > Date: Monday=2C October 1=2C 2012 4:57 PM > To: "user@cassandra.apache.org" > > Subject: Read latency issue >=20 > unning a query to like =93select * from where atag=3D= =94=2C where =91atag=92 is the first column of the composite key=2C from ei= ther JDBC or Hector (equivalent code)=2C results in read times of 200-300ms= from a remote host on the same network. The query returned around 800 resu= lts. Running the same query on a Cassandra host results in a read time of ~= 110-130 ms. > Using read consistency of ONE reduces the read latency by ~20ms=2C compar= ed to using QUORUM. >=20 > Enabling row cache did not seem to change the performance much. Moreover= =2C the row cache =91size=92 according to nodetool was very tiny. Here is a= snapshot of the nodetool info after running few read tests: > Key Cache : size 2448 (bytes)=2C capacity 104857584 (bytes)=2C 231= hits=2C 266 requests=2C 1.000 recent hit rate=2C 14400 save period in seco= nds > Row Cache : size 96 (bytes)=2C capacity 4194304000 (bytes)=2C 9 hi= ts=2C 13 requests=2C NaN recent hit rate=2C 0 save period in seconds >=20 = --_30e76393-e0c6-4a74-b8dc-34347e446a2d_ Content-Type: text/html; charset="Windows-1252" Content-Transfer-Encoding: quoted-printable
Arindam=2C

 =3B Did you also try the cassandra stres= s tool &=3B compare results?

I havent done a pe= rformance test as yet=2C the only ones published on the internet are of YCS= B on an older version of apache cassandra=2C and it doesn't seem to be acti= vely supported or updated
http://www.brianfrankcooper.net/pubs/yc= sb-v4.pdf. =3B

The numbers you have sound very= low=2C for a read of a row by key which should have been the fastest. &nbs= p=3BI hope someone can help investigate or share numbers from their tests.<= /div>

 =3B

Regards=2C
Roshni
 =3B
>=3B From: Dean.Hiller@n= rel.gov
>=3B To: user@cassandra.apache.org
>=3B Date: Tue=2C 2 Oc= t 2012 06:41:09 -0600
>=3B Subject: Re: Read latency issue
>=3B <= br>>=3B Interesting results. With PlayOrm=2C we did a 6 node test of rea= ding 100 rows from 1=2C000=2C000 using PlayOrm Scalable SQL. It only took = 60ms. Maybe we have better hardware though??? We are using 7200 RPM drive= s so nothing fancy on the disk side of things. More nodes puts at a higher= throughput though as reading from more disks will be faster. Anyways=2C y= ou may want to play with more nodes and re-run. If you run a test with Pla= yOrm=2C I would love to know the results there as well.
>=3B
>= =3B Later=2C
>=3B Dean
>=3B
>=3B From: Arindam Barua <=3B= abarua@247-inc.com<=3Bmailto:abarua@247-inc.com>=3B>=3B
>=3B Rep= ly-To: "user@cassandra.apache.org<=3Bmailto:user@cassandra.apache.org>= =3B" <=3Buser@cassandra.apache.org<=3Bmailto:user@cassandra.apache.org&= gt=3B>=3B
>=3B Date: Monday=2C October 1=2C 2012 4:57 PM
>=3B T= o: "user@cassandra.apache.org<=3Bmailto:user@cassandra.apache.org>=3B" = <=3Buser@cassandra.apache.org<=3Bmailto:user@cassandra.apache.org>=3B= >=3B
>=3B Subject: Read latency issue
>=3B
>=3B unning a = query to like =93select * from <=3Btable_name>=3B where atag=3D<=3Bfo= o>=3B=94=2C where =91atag=92 is the first column of the composite key=2C = from either JDBC or Hector (equivalent code)=2C results in read times of 20= 0-300ms from a remote host on the same network. The query returned around 8= 00 results. Running the same query on a Cassandra host results in a read ti= me of ~110-130 ms.
>=3B Using read consistency of ONE reduces the read= latency by ~20ms=2C compared to using QUORUM.
>=3B
>=3B Enablin= g row cache did not seem to change the performance much. Moreover=2C the ro= w cache =91size=92 according to nodetool was very tiny. Here is a snapshot = of the nodetool info after running few read tests:
>=3B Key Cache = : size 2448 (bytes)=2C capacity 104857584 (bytes)=2C 231 hits=2C 266 req= uests=2C 1.000 recent hit rate=2C 14400 save period in seconds
>=3B Ro= w Cache : size 96 (bytes)=2C capacity 4194304000 (bytes)=2C 9 hits= =2C 13 requests=2C NaN recent hit rate=2C 0 save period in seconds
>= =3B
= --_30e76393-e0c6-4a74-b8dc-34347e446a2d_--