From user-return-30307-apmail-cassandra-user-archive=cassandra.apache.org@cassandra.apache.org Tue Nov 27 05:14:37 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 9C233DB87 for ; Tue, 27 Nov 2012 05:14:37 +0000 (UTC) Received: (qmail 42502 invoked by uid 500); 27 Nov 2012 05:14:35 -0000 Delivered-To: apmail-cassandra-user-archive@cassandra.apache.org Received: (qmail 42335 invoked by uid 500); 27 Nov 2012 05:14:34 -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 42292 invoked by uid 99); 27 Nov 2012 05:14:33 -0000 Received: from nike.apache.org (HELO nike.apache.org) (192.87.106.230) by apache.org (qpsmtpd/0.29) with ESMTP; Tue, 27 Nov 2012 05:14:33 +0000 X-ASF-Spam-Status: No, hits=-0.1 required=5.0 tests=HTML_MESSAGE,RCVD_IN_DNSWL_MED,SPF_PASS X-Spam-Check-By: apache.org Received-SPF: pass (nike.apache.org: domain of andras.szerdahelyi@ignitionone.com designates 216.82.249.147 as permitted sender) Received: from [216.82.249.147] (HELO mail29.messagelabs.com) (216.82.249.147) by apache.org (qpsmtpd/0.29) with ESMTP; Tue, 27 Nov 2012 05:14:26 +0000 X-Env-Sender: andras.szerdahelyi@ignitionone.com X-Msg-Ref: server-2.tower-29.messagelabs.com!1353993243!32982411!1 X-Originating-IP: [208.52.173.250] X-StarScan-Received: X-StarScan-Version: 6.6.1.8; banners=-,-,- X-VirusChecked: Checked Received: (qmail 1578 invoked from network); 27 Nov 2012 05:14:04 -0000 Received: from mail.dentsunetwork.com (HELO mail.dentsunetwork.com) (208.52.173.250) by server-2.tower-29.messagelabs.com with AES128-SHA encrypted SMTP; 27 Nov 2012 05:14:04 -0000 Received: from ATL02MB02.corp.local ([fe80::7997:c980:b031:df37]) by ATL02HUB02.corp.local ([::1]) with mapi id 14.02.0318.004; Tue, 27 Nov 2012 00:14:05 -0500 From: Andras Szerdahelyi To: "" Subject: Re: key cache size in mb = 0 doesn't work? Thread-Topic: key cache size in mb = 0 doesn't work? Thread-Index: AQHNzE0wxAEMzwEWX0yijs9cm/6wjJf9ZFsAgAAEcoCAAA5vgA== Date: Tue, 27 Nov 2012 05:14:03 +0000 Message-ID: References: In-Reply-To: Accept-Language: en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [10.0.90.2] Content-Type: multipart/alternative; boundary="_000_FE84AE7AAE9A2B4EA73512EED142BEF830C12CB3ATL02MB02corplo_" MIME-Version: 1.0 X-Virus-Checked: Checked by ClamAV on apache.org --_000_FE84AE7AAE9A2B4EA73512EED142BEF830C12CB3ATL02MB02corplo_ Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: quoted-printable If its the RMI horror ( registry port + actual port ) you are worried about= , enabling MX4J in cassandra-env.sh for the HTTP transport ( http://wiki.ap= ache.org/cassandra/Operations#Monitoring_with_MX4J ) could help although i = never used it myself. Or you can just throw in the Jolokia JVM agent http:/= /www.jolokia.org which will do roughly the same i'm not aware of the hit rate or any of the other cf/system metrics being e= xposed over anything other than JMX, sorry. You can poke around with nodeto= ol on the nodes themselves ( that uses JMX itself ) to get other metrics, b= ut i have not been able to get cache stats out of nodetool cfstats. I think= it might be because they are no longer on the CF's mbean. regards, Andras On 27 Nov 2012, at 05:22, Yiming Sun > wrote: Interesting -- I completely overlooked the OS cache aspect, but how long do= es the OS keep cached pages? It seems they reside in memory for 4 hours at = least (which happens to be the default time for key cache).. The CF has ov= er 2.6 million rows, and I am only fetching a random set of 5000 each time. Where else can I examine the cache hit ratio? The nodes are remote servers = behind some strict university controlled firewall, so it wouldn't be easy f= or me to tap in using JConsole. Thanks. -- Y. On Mon, Nov 26, 2012 at 11:06 PM, Andras Szerdahelyi > wrote: SSTables in http://en.wikipedia.org/wiki/Page_cache maybe? How many rows do you have in this CF? Are you getting all columns? What do the cassandra.db mbeans say ( hit ratio, cache requests, items etc = ) regards, Andras On 27 Nov 2012, at 04:12, Yiming Sun > wrote: > Hi, > > I am carrying out some performance test against a 6-node cassandra cluste= r (v1.1.0), and need to disable the key cache entirely as one of the scenar= ios. > > However, by setting key_cache_size_in_mb to 0, I am still seeing caching = effects. For example, when I fetch a set of 5000 rows, the first time it w= ould deliver at about 60MB/sec, but all subsequent fetches (even after fetc= hing several other sets), it would deliver at about 120MB/sec. This is exa= ctly the same behavior I have observed when the key_cache_size_in_mb was le= ft blank. > > So I am just wondering how I can disable the key cache entirely. All our= nodes have JNA.jar deployed -- not sure it this has anything to do with wh= at I saw. Thanks. > > -- Y --_000_FE84AE7AAE9A2B4EA73512EED142BEF830C12CB3ATL02MB02corplo_ Content-Type: text/html; charset="iso-8859-1" Content-ID: <02C1DD9B1960474D9DCFC08C537A382A@corp.local> Content-Transfer-Encoding: quoted-printable If its the RMI horror ( registry port + actual port ) you are worried a= bout, enabling MX4J in cassandra-env.sh for the HTTP transport ( http://wiki.apache.org/cassandra/Operations#Monitoring_with_MX4J ) co= uld help although i never used it myself. Or you can just throw in the Jolo= kia JVM agent http://www.jolokia.or= g which will do roughly the same

i'm not aware of the hit rate or any of the other cf/system metrics be= ing exposed over anything other than JMX, sorry. You can poke around with n= odetool on the nodes themselves ( that uses JMX itself ) to get other metri= cs, but i have not been able to get cache stats out of nodetool cfstats. I think it might be because they = are no longer on the CF's mbean.

regards,
Andras

On 27 Nov 2012, at 05:22, Yiming Sun <yiming.sun@gmail.com> wrote:

Interesting -- I completely overlooked the OS cac= he aspect, but how long does the OS keep cached pages? It seems they reside= in memory for 4 hours at least (which happens to be the default time for k= ey cache)..  The CF has over 2.6 million rows, and I am only fetching a random set of 5000 each time. 

Where else can I examine the cache hit ratio? The nodes are remote ser= vers behind some strict university controlled firewall, so it wouldn't be e= asy for me to tap in using JConsole.  Thanks.

-- Y.


On Mon, Nov 26, 2012 at 11:06 PM, Andras Szerdah= elyi <andras.szerdahelyi@ignitionone.com> wrote:

SSTables in http://en.wikipedia.org/wiki/Page_cache maybe?
How many rows do you have in this CF? Are you getting all columns?

What do the cassandra.db mbeans say ( hit ratio, cache requests, items etc = )

regards,
Andras

On 27 Nov 2012, at 04:12, Yiming Sun <yiming.sun@gmail.com> wrote:

> Hi,
>
> I am carrying out some performance test against a 6-node cassandra clu= ster (v1.1.0), and need to disable the key cache entirely as one of the sce= narios.
>
> However, by setting key_cache_size_in_mb to 0, I am still seeing cachi= ng effects.  For example, when I fetch a set of 5000 rows, the first t= ime it would deliver at about 60MB/sec, but all subsequent fetches (even af= ter fetching several other sets), it would deliver at about 120MB/sec.  This is exactly the same behavior I have= observed when the key_cache_size_in_mb was left blank.
>
> So I am just wondering how I can disable the key cache entirely.  = ;All our nodes have JNA.jar deployed -- not sure it this has anything to do= with what I saw.  Thanks.
>
> -- Y



--_000_FE84AE7AAE9A2B4EA73512EED142BEF830C12CB3ATL02MB02corplo_--