Mailing-List: contact cassandra-user-help@incubator.apache.org; run by ezmlm
Precedence: bulk
Reply-To: cassandra-user@incubator.apache.org
Received-SPF: neutral (nike.apache.org: local policy)
Content-class: urn:content-classes:message
Subject: Performance of get
MIME-Version: 1.0
Content-Type: multipart/alternative;
	boundary="----_=_NextPart_001_01CA5732.24FBEA9C"
Date: Tue, 27 Oct 2009 11:20:21 -0700
Message-ID: <7FC4BE368A334142B104A2B72F4435AA0452F3@iizuusbs.iizuu.local>
Thread-Topic: Performance of get
Thread-Index: AcpXMiTEfxbCrLtISNCuwsBAx3OuJw==
From: "Christopher McKenzie" <cmckenzie@iizuu.com>
To: <cassandra-user@incubator.apache.org>

This is a multi-part message in MIME format.

------_=_NextPart_001_01CA5732.24FBEA9C
Content-Type: text/plain;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable


Hi,

I'm kind of new to cassandra.  In my application, testing for existence =
is quite usual.  For instance, someone gives me say, a URL, and if it =
doesn't exist, I am ok with that.  This will be the usual case.  =
Existence may only happen like 20% of the time.

I was thinking this would be fine, until I saw that the thrift API =
*throws and exception* if a key doesn't exist.

Checking for existence, a normal control flow operation, requires using =
a try/catch method, which are not intended for normal control flow. Many =
languages discard the stack when exceptions are thrown and then do =
non-local jumps.  Other languages will throw the exception to the =
operating system (a 'first chance' exception for instance) to be caught =
by a debugger if attached.  Other languages actually execute an =
interrupt on the processor which is a horrible design principle for =
normal program flow: http://en.wikipedia.org/wiki/Interrupt.

My experience with PHP has shown that interrupts come at a cost of about =
200 times that of a type check.  C++ if I recall, on x86/linux was at =
least an order of magnitude as well.  In Java, it's a bit cheaper, but =
there are still faster ways of doing it.

Could Get possibly return a reserved value instead?

Thanks.

~chris.

------_=_NextPart_001_01CA5732.24FBEA9C
Content-Type: text/html;
	charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 3.2//EN">
<HTML>
<HEAD>
<META HTTP-EQUIV=3D"Content-Type" CONTENT=3D"text/html; =
charset=3Diso-8859-1">
<META NAME=3D"Generator" CONTENT=3D"MS Exchange Server version =
6.5.7654.12">
<TITLE>Performance of get</TITLE>
</HEAD>
<BODY>
<!-- Converted from text/plain format -->
<BR>

<P><FONT SIZE=3D2>Hi,<BR>
<BR>
I'm kind of new to cassandra.&nbsp; In my application, testing for =
existence is quite usual.&nbsp; For instance, someone gives me say, a =
URL, and if it doesn't exist, I am ok with that.&nbsp; This will be the =
usual case.&nbsp; Existence may only happen like 20% of the time.<BR>
<BR>
I was thinking this would be fine, until I saw that the thrift API =
*throws and exception* if a key doesn't exist.<BR>
<BR>
Checking for existence, a normal control flow operation, requires using =
a try/catch method, which are not intended for normal control flow. Many =
languages discard the stack when exceptions are thrown and then do =
non-local jumps.&nbsp; Other languages will throw the exception to the =
operating system (a 'first chance' exception for instance) to be caught =
by a debugger if attached.&nbsp; Other languages actually execute an =
interrupt on the processor which is a horrible design principle for =
normal program flow: <A =
HREF=3D"http://en.wikipedia.org/wiki/Interrupt">http://en.wikipedia.org/w=
iki/Interrupt</A>.<BR>
<BR>
My experience with PHP has shown that interrupts come at a cost of about =
200 times that of a type check.&nbsp; C++ if I recall, on x86/linux was =
at least an order of magnitude as well.&nbsp; In Java, it's a bit =
cheaper, but there are still faster ways of doing it.<BR>
<BR>
Could Get possibly return a reserved value instead?<BR>
<BR>
Thanks.<BR>
<BR>
~chris.</FONT>
</P>

</BODY>
</HTML>
------_=_NextPart_001_01CA5732.24FBEA9C--