Re: load balancer models and performance

Right - I agree that an ADC should be "green" and as energy-efficient as
possible - when it comes to data center space/cooling costs (i.e. $$$),
one needs to see it complies with all requirements. 

But here's another thought: I'm not sure that the need for a
fully-equipped ADC (what described below as an 'advanced ADC' platform -
and I agree in this respect) contradicts incremental throughput
licensing / 'pay as you grow' approach (it has several names out
there...).  On the contrary, I actually think they can go together
pretty well: I see the 0-4Gbps throughput market segment as a very
dynamic one - throughput-wise. Therefore, in my philosophy, if you could
pay exactly for what you need today (and in case you'll need more
capacity, you'll pay only for the delta) --  while getting a
fully-equipped ADC upfront regardless of what you need today -- you can
ensure full flexibility and enjoy both worlds.   


Nir Ilani
Director, Advanced ADC, Radware


-----Original Message-----
From: [email protected] [mailto:[email protected]] On Behalf
Of Tal Klein
Sent: Friday, July 24, 2009 8:39 PM
To: Load Balancing Mailing List
Subject: Re: [load balancing] load balancer models and performance

Actually Ed, I'd like to offer a counterpoint to "At the end of the day,
the hardware is a commodity and a small part of the cost of delivering
the appliance" -- We actually have lots good reasons for different
hardware platforms which have to do with all the extra stuff we throw
into our hardware platforms (processors, ram, asics, ports, etc.), but
there is also an important customer consideration here: The
power+cooling costs which some of our customers are very sensitive to.
For example, some of our larger internet customers determine what device
they purchase based on a BTU per rack calculation, so sometimes that
means having to make a compromise between throughput and power+cooling
requirements. This means that although in theory they could have one
device that handles multiple racks' worth of bandwidth, sometimes it
makes more sense for them to have a dedicated lower end device per rack.

Food for thought :)

-Tal
---
Tal M. Klein
Integrated Product Strategy
Citrix Systems, Inc.

________________________________________
From: [email protected] [[email protected]] On Behalf Of Ed
Toro [[email protected]]
Sent: Friday, July 24, 2009 9:05 AM
To: Load Balancing Mailing List
Subject: Re: [load balancing] load balancer models and performance

My apologies for suggesting F5 capped raw throughput. I was probably
thinking of compression and SSL. I'm not sure if Cisco or NetScaler do
this. If not, I bet they are looking at it. At the end of the day, the
hardware is a commodity and a small part of the cost of delivering the
appliance. It makes more sense to sell the customer a big appliance they
can upgrade performance on easily for more revenue down the road. If
they have to rip out a physical appliance then they are less likely to
exert the effort. It also opens the door to other vendors if they have
to do a physical replacement.


Kirk Bauer wrote:
> The only two items where performance is limited by license on a BIG-IP
is total compressed throughput (you get 50Mbps and can unlock more) and
SSL transactions per second (you get 500 and can unlock more).
>

_______________________________________________
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
_______________________________________________
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
_______________________________________________
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