From mboxrd@z Thu Jan 1 00:00:00 1970 Received: from eggs.gnu.org ([140.186.70.92]:32878) by lists.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1S0Tyk-0007v9-IW for qemu-devel@nongnu.org; Thu, 23 Feb 2012 03:24:48 -0500 Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71) (envelope-from ) id 1S0Tye-0006VM-V7 for qemu-devel@nongnu.org; Thu, 23 Feb 2012 03:24:42 -0500 Received: from mail-lpp01m010-f45.google.com ([209.85.215.45]:39876) by eggs.gnu.org with esmtp (Exim 4.71) (envelope-from ) id 1S0Tye-0006V9-LM for qemu-devel@nongnu.org; Thu, 23 Feb 2012 03:24:36 -0500 Received: by lahi5 with SMTP id i5so1236549lah.4 for ; Thu, 23 Feb 2012 00:24:35 -0800 (PST) MIME-Version: 1.0 From: PANKAJ RAWAT Date: Thu, 23 Feb 2012 13:54:15 +0530 Message-ID: Content-Type: multipart/alternative; boundary=f46d043086c0cfba5204b99d5f18 Subject: [Qemu-devel] effect of cluster_size on qcow2 performance during expansion List-Id: List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , To: qemu-devel --f46d043086c0cfba5204b99d5f18 Content-Type: text/plain; charset=ISO-8859-1 Hi all I have been working on the qcow2 image format , In theory regarding cluster size it is written that as the size of cluster increase performance should increase. But something surprising happen The performance is degrading as the size of cluster increased from *64K* to *1M* or *2M* ( during expansion of qcow2 image) By expansion i mean that initially qcow2 image is small in size and size increases when we perform any write operation on it. Here are the reading taken through fio tool during write operation *Cluster size : 32K 587Mb/sec (aggrb MB/sec) * * * *Cluster size : 64K 697Mb/sec** (aggrb MB/sec)* * Cluster size : 1M 440Mb/sec* *(aggrb MB/sec)* * * * Cluster size : 2M 289Mb/sec** (aggrb MB/sec)* as in above it can bee seen that the 64k is performing better can any one tell why ? -- *Pankaj Rawat* --f46d043086c0cfba5204b99d5f18 Content-Type: text/html; charset=ISO-8859-1 Content-Transfer-Encoding: quoted-printable

Hi a= ll

I have been working on= the qcow2 image format ,

In theory regarding = =A0cluster size it is written that as the size of cluster increase performance should increase.

=A0

But something surprisi= ng happen The performance is degrading as the size of cluster increased from 64K t= o 1M or 2M=A0 ( during expansion of qcow2 image)

=A0

<= p class=3D"MsoNormal">By expansion i m= ean that initially qcow2 image is small in size and size increases when we = perform any write operation on it.


=

Here= are the reading taken through fio tool during write operation


= Cluster size : 32K =A0=A0 587Mb/sec (aggrb MB/sec)

Cluster size : 64K =A0=A0 697Mb/sec (aggrb MB/sec)

Cluster size : 1M =A0 =A0 440Mb/sec=A0 (aggrb MB/sec)


<= /b>

Cluster size : 2M =A0 =A0 289Mb/sec=A0 (aggrb MB/sec)

as = in above it can bee seen that the 64k is performing better

can any one tel= l why ?




--
Pankaj Rawat

--f46d043086c0cfba5204b99d5f18--