X-MimeOLE: Produced By Microsoft Exchange V6.5
Received: by onstor-exch02.onstor.net 
	id <01C78034.FE06FF6F@onstor-exch02.onstor.net>; Mon, 16 Apr 2007 07:39:09 -0700
MIME-Version: 1.0
Content-Type: text/plain;
	charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Content-class: urn:content-classes:message
Subject: RE: Measurement of Linux Small File Performance
Date: Mon, 16 Apr 2007 07:39:09 -0700
Message-ID: <BB375AF679D4A34E9CA8DFA650E2B04E03436C0A@onstor-exch02.onstor.net>
In-Reply-To: <BB375AF679D4A34E9CA8DFA650E2B04E0338F9C9@onstor-exch02.onstor.net>
X-MS-Has-Attach: 
X-MS-TNEF-Correlator: 
Thread-Topic: Measurement of Linux Small File Performance
Thread-Index: Acd9R+/CtNYa2peJQGiYwBvvv2t1aAC7Mu2Q
From: "Raj Kumar" <raj.kumar@onstor.com>
To: "Jay Michlin" <jay.michlin@onstor.com>,
	"dl-Delorean" <dl-Delorean@onstor.com>
Cc: "Maxim Kozlovsky" <maxim.kozlovsky@onstor.com>

Since Linux cp performance issue will not be addressed in Delorean, I
will request Sandrine to schedule to do this test post Delorean.

All the data sets and client information are documented for the future
tests.

Thanks.

-----Original Message-----
From: Jay Michlin=20
Sent: Thursday, April 12, 2007 2:17 PM
To: dl-Delorean
Cc: Maxim Kozlovsky
Subject: Measurement of Linux Small File Performance

In today's Delorean team meeting the topic of measuring small file
performance was discussed.

The Windows part of the measurement was clarified to require an
apples-to-apples comparison of identical hardware, configurations and
workloads in the lab. The only difference between the two configurations
to be compared is that one will have a Bobcat and the other a Windows
server. This will give us a reliable quantification of our performance
position versus Windows Explorer. John said he can create the required
environment in the lab, and Raj agreed that QA will run this test.

The question arose of whether to run a similar comparison for Linux cp
performance. Since there is no expectation of further development on
this for Delorean, it may be unnecessary. However, I asked that we check
with Max as developer of the code before making a firm decision not to
run such a test. Raj was reluctant to check, so I agreed to.

Max observed that we need an apples-to-apples measure of baseline Linux
cp performance so we can assess any additional development we do,
whether for Delorean of future projects. If the QA schedule can't
support this before Delorean FCS, then it should be scheduled
immediately after. However, since the lab will be set up to do this test
for Windows, it would be more efficient to do earlier.

It occurred to me also that whatever our small file performance versus
Linux may be in Delorean, we probably ought to know what it is for
marketing and selling purposes.

So I ask again for this Linux apples-to-apples measure, before Delorean
FCS if at all possible. If this is difficult or if there is
disagreement, we ought to reraise the matter at the next Delorean
meeting.

