Robert Parkhurst | 2 Jul 23:48 2007
Picon

network switch benchmark comparision

I'm looking into network switches at work.  We're standardized on D-Link GbE "workgroup" switches, but I don't think we're getting a lot out of them (i.e. general LAN use seems to go slower than it should being GbE).  I'm curious if anyone knows of any sort of benchmark comparisions between various switches (like D-Link, Linksys, Cisco, Extreme, Netgear, etc.).


Thanks

Robert

<div><p>I'm looking into network switches at work.&nbsp; We're standardized on D-Link GbE "workgroup" switches, but I don't think we're getting a lot out of them (i.e. general LAN use seems to go slower than it should being GbE).&nbsp; I'm curious if anyone knows of any sort of benchmark comparisions between various switches (like D-Link, Linksys, Cisco, Extreme, Netgear, etc.).
<br><br><br>Thanks<br><br>Robert<br></p></div>
Travis | 3 Jul 01:51 2007

Re: network switch benchmark comparision

On Mon, Jul 02, 2007 at 04:48:46PM -0500, Robert Parkhurst wrote:
> I'm looking into network switches at work.  We're standardized on D-Link GbE
> "workgroup" switches, but I don't think we're getting a lot out of them 
> (i.e.
> general LAN use seems to go slower than it should being GbE).  I'm curious
> if anyone knows of any sort of benchmark comparisions between various
> switches (like D-Link, Linksys, Cisco, Extreme, Netgear, etc.).

Ultimately, you have two things to really worry about:  how fast/large the
switch fabric is on the backplane and how many uplinks you have between
your switches to your cores.

From the descriptions of the D-link hardware that I see on their website, it
looks like their switch fabric is capable of driving all end-user ports at full
duplex at the same time (so, 24 Gb ports has a 48Gbps switch fabric).  I can't
tell if their uplink ports can be used in addition to all ports or if they
"piggyback" on some of the copper ports (for example, I know Nortel baystacks
treat ports 47 and 48 as either copper or fiber uplinks, but not both
at the same time).

I think the real issue you're going to be dealing with is the number of
uplinks you're sticking on each switch and being able to trunk your uplinks
together for increased bandwidth.  One thing to realize is that when you're
doing trunking like this, any one system contacting another system on the
other side of the uplink is still only going to get a maximum of 1Gb/s.  
Link aggregation gives you the ability to expand bandwidth overall so you
can have more concurrent data streams.  (eg, a 4Gb trunk can have four 1Gb
streams going, not one 4Gb stream).

BTW, I was looking at the "Web Smart" line of Gigabite D-link switches when
I responded to this, just so you have a reference to what I was looking at.

Probably doesn't help answer your question directly, but gives you an idea
of what to look for in a switch and where your bottlenecks are going to
appear first.

Travis
-- 
Travis Campbell
hcoyote@...
On Mon, Jul 02, 2007 at 04:48:46PM -0500, Robert Parkhurst wrote:
> I'm looking into network switches at work.  We're standardized on D-Link GbE
> "workgroup" switches, but I don't think we're getting a lot out of them 
> (i.e.
> general LAN use seems to go slower than it should being GbE).  I'm curious
> if anyone knows of any sort of benchmark comparisions between various
> switches (like D-Link, Linksys, Cisco, Extreme, Netgear, etc.).

Ultimately, you have two things to really worry about:  how fast/large the
switch fabric is on the backplane and how many uplinks you have between
your switches to your cores.

From the descriptions of the D-link hardware that I see on their website, it
looks like their switch fabric is capable of driving all end-user ports at full
duplex at the same time (so, 24 Gb ports has a 48Gbps switch fabric).  I can't
tell if their uplink ports can be used in addition to all ports or if they
"piggyback" on some of the copper ports (for example, I know Nortel baystacks
treat ports 47 and 48 as either copper or fiber uplinks, but not both
at the same time).

I think the real issue you're going to be dealing with is the number of
uplinks you're sticking on each switch and being able to trunk your uplinks
together for increased bandwidth.  One thing to realize is that when you're
doing trunking like this, any one system contacting another system on the
other side of the uplink is still only going to get a maximum of 1Gb/s.  
Link aggregation gives you the ability to expand bandwidth overall so you
can have more concurrent data streams.  (eg, a 4Gb trunk can have four 1Gb
streams going, not one 4Gb stream).

BTW, I was looking at the "Web Smart" line of Gigabite D-link switches when
I responded to this, just so you have a reference to what I was looking at.

Probably doesn't help answer your question directly, but gives you an idea
of what to look for in a switch and where your bottlenecks are going to
appear first.

Travis
--

-- 
Travis Campbell
hcoyote@...
Mark R Ward | 3 Jul 04:11 2007

Mark R Ward/GIS/CSC is out of the office.


I will be out of the office starting  07/01/2007 and will not return until
07/09/2007.

I will be out the week of Jully 4th. Please contact Sly Bernal in my
absence if you need help.

Jim C. Nasby | 3 Jul 17:48 2007

Re: network switch benchmark comparision

On a related note... anyone know of any comparisons of lower-end Gig-E
switches? I bought one recently and tried to find a good one, but the
latency seems really crappy (> 1ms, unloaded).

On Mon, Jul 02, 2007 at 06:51:51PM -0500, Travis wrote:
> On Mon, Jul 02, 2007 at 04:48:46PM -0500, Robert Parkhurst wrote:
> > I'm looking into network switches at work.  We're standardized on D-Link GbE
> > "workgroup" switches, but I don't think we're getting a lot out of them 
> > (i.e.
> > general LAN use seems to go slower than it should being GbE).  I'm curious
> > if anyone knows of any sort of benchmark comparisions between various
> > switches (like D-Link, Linksys, Cisco, Extreme, Netgear, etc.).
> 
> Ultimately, you have two things to really worry about:  how fast/large the
> switch fabric is on the backplane and how many uplinks you have between
> your switches to your cores.
> 
> From the descriptions of the D-link hardware that I see on their website, it
> looks like their switch fabric is capable of driving all end-user ports at full
> duplex at the same time (so, 24 Gb ports has a 48Gbps switch fabric).  I can't
> tell if their uplink ports can be used in addition to all ports or if they
> "piggyback" on some of the copper ports (for example, I know Nortel baystacks
> treat ports 47 and 48 as either copper or fiber uplinks, but not both
> at the same time).
> 
> I think the real issue you're going to be dealing with is the number of
> uplinks you're sticking on each switch and being able to trunk your uplinks
> together for increased bandwidth.  One thing to realize is that when you're
> doing trunking like this, any one system contacting another system on the
> other side of the uplink is still only going to get a maximum of 1Gb/s.  
> Link aggregation gives you the ability to expand bandwidth overall so you
> can have more concurrent data streams.  (eg, a 4Gb trunk can have four 1Gb
> streams going, not one 4Gb stream).
> 
> BTW, I was looking at the "Web Smart" line of Gigabite D-link switches when
> I responded to this, just so you have a reference to what I was looking at.
> 
> 
> Probably doesn't help answer your question directly, but gives you an idea
> of what to look for in a switch and where your bottlenecks are going to
> appear first.
> 
> 
> Travis
> -- 
> Travis Campbell
> hcoyote@...

> _______________________________________________
> ALG-technical mailing list http://austinlug.org/cgi-bin/mailman/listinfo/alg-technical

-- 
Decibel!, aka Jim C. Nasby, Database Architect  decibel@... 
Give your computer some brain candy! www.distributed.net Team #1828
On a related note... anyone know of any comparisons of lower-end Gig-E
switches? I bought one recently and tried to find a good one, but the
latency seems really crappy (> 1ms, unloaded).

On Mon, Jul 02, 2007 at 06:51:51PM -0500, Travis wrote:
> On Mon, Jul 02, 2007 at 04:48:46PM -0500, Robert Parkhurst wrote:
> > I'm looking into network switches at work.  We're standardized on D-Link GbE
> > "workgroup" switches, but I don't think we're getting a lot out of them 
> > (i.e.
> > general LAN use seems to go slower than it should being GbE).  I'm curious
> > if anyone knows of any sort of benchmark comparisions between various
> > switches (like D-Link, Linksys, Cisco, Extreme, Netgear, etc.).
> 
> Ultimately, you have two things to really worry about:  how fast/large the
> switch fabric is on the backplane and how many uplinks you have between
> your switches to your cores.
> 
> From the descriptions of the D-link hardware that I see on their website, it
> looks like their switch fabric is capable of driving all end-user ports at full
> duplex at the same time (so, 24 Gb ports has a 48Gbps switch fabric).  I can't
> tell if their uplink ports can be used in addition to all ports or if they
> "piggyback" on some of the copper ports (for example, I know Nortel baystacks
> treat ports 47 and 48 as either copper or fiber uplinks, but not both
> at the same time).
> 
> I think the real issue you're going to be dealing with is the number of
> uplinks you're sticking on each switch and being able to trunk your uplinks
> together for increased bandwidth.  One thing to realize is that when you're
> doing trunking like this, any one system contacting another system on the
> other side of the uplink is still only going to get a maximum of 1Gb/s.  
> Link aggregation gives you the ability to expand bandwidth overall so you
> can have more concurrent data streams.  (eg, a 4Gb trunk can have four 1Gb
> streams going, not one 4Gb stream).
> 
> BTW, I was looking at the "Web Smart" line of Gigabite D-link switches when
> I responded to this, just so you have a reference to what I was looking at.
> 
> 
> Probably doesn't help answer your question directly, but gives you an idea
> of what to look for in a switch and where your bottlenecks are going to
> appear first.
> 
> 
> Travis
> -- 
> Travis Campbell
> hcoyote@...

> _______________________________________________
> ALG-technical mailing list http://austinlug.org/cgi-bin/mailman/listinfo/alg-technical

--

-- 
Decibel!, aka Jim C. Nasby, Database Architect  decibel@... 
Give your computer some brain candy! www.distributed.net Team #1828
Jim Parkhurst | 3 Jul 17:54 2007
Picon
Picon

Re: Mark R Ward/GIS/CSC is out of the office.

Don't you just "love" folks who set "Reply-To-ALL" in their emailers?  :-|

>>> Mark R Ward <mward31@...> 07/02/2007 21:11 >>>

I will be out of the office starting  07/01/2007 and will not return until
07/09/2007.

I will be out the week of Jully 4th. Please contact Sly Bernal in my
absence if you need help.

_______________________________________________
ALG-technical mailing list http://austinlug.org/cgi-bin/mailman/listinfo/alg-technical

Tony Jarnigan | 3 Jul 18:00 2007
Picon

Re: network switch benchmark comparision

On 7/3/07, Jim C. Nasby <decibel-ft4DTYm7hStAfugRpC6u6w@public.gmane.org> wrote:

On a related note... anyone know of any comparisons of lower-end Gig-E
switches? I bought one recently and tried to find a good one, but the
latency seems really crappy (> 1ms, unloaded).

Try this:

http://www.smallnetbuilder.com/content/view/29915/51/
<div>
<p>On 7/3/07, Jim C. Nasby &lt;<a href="mailto:decibel@...">decibel@...</a>&gt; wrote:</p>
<div>
<span class="gmail_quote"></span><blockquote class="gmail_quote">
On a related note... anyone know of any comparisons of lower-end Gig-E<br>switches? I bought one recently and tried to find a good one, but the<br>latency seems really crappy (&gt; 1ms, unloaded).<br>
</blockquote>
</div>
<br>
Try this:<br><br><a href="http://www.smallnetbuilder.com/content/view/29915/51/">http://www.smallnetbuilder.com/content/view/29915/51/</a><br>
</div>
Jim C. Nasby | 3 Jul 21:33 2007

Re: network switch benchmark comparision

On Tue, Jul 03, 2007 at 11:00:38AM -0500, Tony Jarnigan wrote:
> On 7/3/07, Jim C. Nasby <decibel@...> wrote:
> >
> >On a related note... anyone know of any comparisons of lower-end Gig-E
> >switches? I bought one recently and tried to find a good one, but the
> >latency seems really crappy (> 1ms, unloaded).
> >
> 
> Try this:
> 
> http://www.smallnetbuilder.com/content/view/29915/51/

Thanks, I'll have to keep that site in mind next time I'm buying
infrastructure!
-- 
Decibel!, aka Jim C. Nasby, Database Architect  decibel@... 
Give your computer some brain candy! www.distributed.net Team #1828
On Tue, Jul 03, 2007 at 11:00:38AM -0500, Tony Jarnigan wrote:
> On 7/3/07, Jim C. Nasby <decibel@...> wrote:
> >
> >On a related note... anyone know of any comparisons of lower-end Gig-E
> >switches? I bought one recently and tried to find a good one, but the
> >latency seems really crappy (> 1ms, unloaded).
> >
> 
> Try this:
> 
> http://www.smallnetbuilder.com/content/view/29915/51/

Thanks, I'll have to keep that site in mind next time I'm buying
infrastructure!
--

-- 
Decibel!, aka Jim C. Nasby, Database Architect  decibel@... 
Give your computer some brain candy! www.distributed.net Team #1828
Thomas Cameron | 3 Jul 22:10 2007

Re: network switch benchmark comparision

Jim C. Nasby wrote:
> On Tue, Jul 03, 2007 at 11:00:38AM -0500, Tony Jarnigan wrote:
>> On 7/3/07, Jim C. Nasby <decibel@...> wrote:
>>> On a related note... anyone know of any comparisons of lower-end Gig-E
>>> switches? I bought one recently and tried to find a good one, but the
>>> latency seems really crappy (> 1ms, unloaded).
>>>
>> Try this:
>>
>> http://www.smallnetbuilder.com/content/view/29915/51/
> 
> Thanks, I'll have to keep that site in mind next time I'm buying
> infrastructure!

FWIW I have a Linksys JGS524 switch and I love it.  24 ports, 48 
gig/second backplane, it's nice.

I have not benchmarked it per se, but I know that when I moved from the 
10/100 to the 10/100/1000 I said "holy &$ <at> %!" the first time I copied 
ISOs across the network.

Also, my kickstarts are waaaaay faster now.

TC
Donn Washburn | 7 Jul 02:16 2007
Picon

Compile error I see a lot

Hwy Group;

Not being a programmer I have a question.
This error normally is just a warning but I am curious what it means.

If it something I could fix I would do it
-- 
73 de Donn Washburn
307 Savoy Street     Email: " n5xwb@... "
Sugar Land, TX 77478 LL# 1.281.242.3256
Ham Callsign N5XWB   HAMs : " n5xwb@... "
VoIP via Gizmo: bmw_87kbike / via Skype: n5xwbg
BMW MOA #: 4146 - Ambassador
       " http://counter.li.org " #279316

Did you know?
The transistor was invented by three white men.
  Fl_RGB_Image(const uchar *bits, int W, int H, int D=3, int LD=0) :
    Fl_Image(W,H,D), array(bits), alloc_array(0), id(0), mask(0) {data((const char **)&array, 1); ld(LD);}

error 

dereferencing type-punned pointer will break strict-aliasing rules
  Fl_RGB_Image(const uchar *bits, int W, int H, int D=3, int LD=0) :
    Fl_Image(W,H,D), array(bits), alloc_array(0), id(0), mask(0) {data((const char **)&array, 1); ld(LD);}

error 

dereferencing type-punned pointer will break strict-aliasing rules
John Doe | 7 Jul 05:42 2007
Picon

Re: Compile error I see a lot

Do a man on gcc, and read the section on "-fstrict-aliasing".  It describes what type-punning is and when it is considered valid in the presence of the optimization.  I'd also email whoever the package maintainer is, and see if they can fix the warning for you.

Juan

On 7/6/07, Donn Washburn <n5xwb-xV8QqCHEK+cdnm+yROfE0A@public.gmane.org> wrote:
Hwy Group;

Not being a programmer I have a question.
This error normally is just a warning but I am curious what it means.

If it something I could fix I would do it
--
73 de Donn Washburn
307 Savoy Street     Email: " n5xwb-xV8QqCHEK+cdnm+yROfE0A@public.gmane.org "
Sugar Land, TX 77478 LL# 1.281.242.3256
Ham Callsign N5XWB   HAMs : " n5xwb-WYrOkVUspZo@public.gmane.org "
VoIP via Gizmo: bmw_87kbike / via Skype: n5xwbg
BMW MOA #: 4146 - Ambassador
       " http://counter.li.org " #279316

Did you know?
The transistor was invented by three white men.

  Fl_RGB_Image(const uchar *bits, int W, int H, int D=3, int LD=0) :
    Fl_Image(W,H,D), array(bits), alloc_array(0), id(0), mask(0) {data((const char **)&array, 1); ld(LD);}

error

dereferencing type-punned pointer will break strict-aliasing rules

_______________________________________________
ALG-technical mailing list http://austinlug.org/cgi-bin/mailman/listinfo/alg-technical


<div>
<p>Do a man on gcc, and read the section on "-fstrict-aliasing".&nbsp; It describes what type-punning is and when it is considered valid in the presence of the optimization.&nbsp; I'd also email whoever the package maintainer is, and see if they can fix the warning for you.
<br><br>Juan<br><br></p>
<div>
<span class="gmail_quote">On 7/6/07, Donn Washburn &lt;<a href="mailto:n5xwb@...">n5xwb@...</a>&gt; wrote:</span><blockquote class="gmail_quote">
Hwy Group;<br><br>Not being a programmer I have a question.<br>This error normally is just a warning but I am curious what it means.<br><br>If it something I could fix I would do it<br>--<br>73 de Donn Washburn<br>307 Savoy Street&nbsp;&nbsp;&nbsp;&nbsp; Email: " 
<a href="mailto:n5xwb@...">n5xwb@...</a> "<br>Sugar Land, TX 77478 LL# 1.281.242.3256<br>Ham Callsign N5XWB&nbsp;&nbsp; HAMs : " <a href="mailto:n5xwb@...">n5xwb@...</a> "<br>VoIP via Gizmo: bmw_87kbike / via Skype: n5xwbg
<br>BMW MOA #: 4146 - Ambassador<br>&nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; " <a href="http://counter.li.org">http://counter.li.org</a> " #279316<br><br>Did you know?<br>The transistor was invented by three white men.<br><br>&nbsp;&nbsp;Fl_RGB_Image(const uchar *bits, int W, int H, int D=3, int LD=0) :
<br>&nbsp;&nbsp;&nbsp;&nbsp;Fl_Image(W,H,D), array(bits), alloc_array(0), id(0), mask(0) {data((const char **)&amp;array, 1); ld(LD);}<br><br>error<br><br>dereferencing type-punned pointer will break strict-aliasing rules<br><br>_______________________________________________
<br>ALG-technical mailing list <a href="http://austinlug.org/cgi-bin/mailman/listinfo/alg-technical">http://austinlug.org/cgi-bin/mailman/listinfo/alg-technical</a><br><br>
</blockquote>
</div>
<br>
</div>

Gmane