Re: load balancer models and performance

Andre,

 

It's early Seattle time, so I can't validate this with product management,
but . . .

 

The difference in the performance, as I remember it, was because of a
software upgrade (I'm not 100% sure which software version included this-it
looks like v9.4).  Good, bad or otherwise, F5 has almost always taken a
hybrid approach--meaning that our software and hardware are very tightly
coupled; we believe it is the best way.  Changes in either one often result
in significant improvement in overall performance.  When the new 1600
hardware was brought out, it performed 50% better than the 1500 platform
running the same software (750Mpbs versus 500Mbps) because of the increased
capability of the raw hardware.  When the following software release came
out, adjustments in the software (flat out improvements in our OS and
specific tweaks to take better advantage of the new hardware) improved
performance again-to the point that we republished the spec sheet to reflect
the new 1Gbps.  I want to say that happened near the end of last summer, but
again, I'm not 100% sure on the when.

 

Back to your real need, however, the 3600 sounds like right size range given
that the 1600, even with the new specs, is on the low-end of your need, but,
as I've said 100 times on this list-take our published numbers or the
published numbers of any vendor with a grain of salt.  These numbers are
always best-case in pristine LANs with very little if anything else going
on.  As much as we all know that it isn't any reflection of the real-world,
we also cannot possible emulate all the circumstance we would need to give
numbers you could really use.  BTW-this goes for all those "performance
reports" that many vendors get third-party 'testing' companies to do.  Just
ONCE, I'd like to see a report come out where the company paying for it
DIDN'T win the evaluation; you'd think that statistically, if all the
testing and reporting was done above board, it would have to happen sooner
or later.  I've even been contacted by one of these companies who was
completing a test for a competitor and they wanted me to know that we were
actually cleaning the competitors clock no matter what they did-then wanted
to know how much I'd pay for the data and a report (I'm happy to say, we
declined to participate).

 

My Point?  These numbers are highly generalized guidelines-the only way to
really know is to put the box in your network with your applications and
your users and find out what the scoop is.  Don't 'trust' a single one of us
vendors-bring us ALL in and make us satisfy your base requirements and THEN
look at any additional features/functions you may or may not need down the
road.  If we can't satisfy your base needs-all the rest is academic.  Use
the published numbers to find the right ballpark of device (obviously, if
the vendor gives a best-case of 500Mbps, it probably won't do 1.5Gbps).  The
time and money spent to really examine your options will be made up for a
thousand fold by ensuring that you select the best product FOR YOUR
ORGANIZATION.

 

I hope that helps!!!

 


KJ (Ken) Salchow, Jr. | Manager, Technical Marketing


D 651.423.1133

M 612.868.1258

P 206.272.5555

F 206.272.5555

 <http://www.f5.com/> www.f5.com

 

 

From: [email protected] [mailto:[email protected]] On Behalf Of
Andrejus Baranovskis
Sent: Friday, July 24, 2009 3:26 AM
To: [email protected]
Subject: [load balancing] load balancer models and performance

 

Hello list,

 

Looking at load balancer options to purchase for our medium sized company.
We've been looking at F5, Netscaler and  Cisco.

 

We were looking for a capacity of 1-2 Gbps SSL/load balance. We got a little
confused with F5 models.  1600 series was out of question as it's datasheet
used to say 750 Mbps and we were looking at 3600 or 6900.

 

However there are new datasheets now which say 1600 is 1 Gbps a nd 3600 is 2
Gbps (used to be 1.5 Gbps). Naturally we are confused. Does anyone have
clear numbers for this? Why did the specs change all of a sudden? any change
on the CPU/HW? As far as I know these are new models which came out very
recently and there is no recent major software release in the devcentral to
justify the improvements in the datasheets.

 

Thoughts/opinions? If anyone has field performance experience with these
models it will be very valuable for us, we don't have much time to test
every vendor and each model.

 

Andre

_______________________________________________
lb-l mailing list
[email protected]
http://vegan.net/mailman/listinfo/lb-l
Searchable Archive: http://vegan.net/lb/archive
http://lbdigest.com Load Balancing Digest
http://lbwiki.com Load Balancing Wiki