Re: Zeus ZXTM?

Kenneth Salchow <[email protected]> Wed, 8 Jul 2009 09:44:20 -0700
Newsgroups gmane.comp.programming.load-balancing.general
Message-ID <4B18A8F75A6384449755BC7784073E9360CF4C448D@exch11.olympus.f5net.com>
Wesley,

 

Thanks for the kudos as well as the humbling.  J

 

Yeah, one of the big changes between v4 and v9 was that v9 *is* a
full-proxy—period.  So, with v4 there were a lot of things we did naturally
and you had to specifically configure it to act like it would proxy some
connections.  In v9, we naturally proxy and you have to specifically
configure it to pretend not too—if that makes any sense. 

 

Having been a customer before becoming an employee—I know exactly what you
mean from both a feature standpoint as well as a support standpoint.  I used
to say that F5 consistently provides its customers with enough rope to hang
themselves with; and have dozens of stories of how customers (including
myself) did just that.  Like the company that created elaborate EAV checks
on their application that took about 45 seconds to complete—and scheduled
them to run every 15 seconds; I’d never seen a fully loaded catalyst switch
with all solid lights before that night.  Or setting your system to validate
outbound routes and picking targets on the Internet you don’t own, who then
think you’re attacking them, etc.  I used to also joke about waiting for the
IP police to show up and haul us away for the things we did to IP packets
with the BIG-IP.

 

On the support front—I hope that finally getting an answer has more to do
with the concerted effort we have made over the past several years to
improve our support capability including numbers of engineers, training of
engineers and improved escalation procedures, than it was just finally
getting a really good engineer.  It’s been something we have invested
significantly into and all the metrics and reports I’ve seen show that it
has really paid off and improved our support significantly.  I know it is
head and shoulders above the old, old days when I (at the time an MCSE whose
only Unix experience was DOS) seemed to know more about BSD than some of the
people on the phone.  Or when I explained how I managed to load-balance DCOM
by enabling virtual-servers AND IP forwarding simultaneously and was told it
was impossible to do that. 

 

Ah, the good old days.  Bright-Tiger Cluster-Cats anyone?

 

 


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


D 651.423.1133

M 612.868.12588

P 206.272.5555

F 206.272.5555

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

 

 

From: [email protected] [mailto:[email protected]] On Behalf Of
Wesley North
Sent: Wednesday, July 08, 2009 11:06 AM
To: Load Balancing Mailing List
Subject: Re: [load balancing] Zeus ZXTM?

 

F5 is a great product, if one could understand just how many bells and
whistles there are and what happens when you turn one of them. Support is a
major issue. In the V4 days turning on IP forwarding for an inline appliance
allowed stateless IP forwarding, with v9 that is not available out of the
box. So we create virtual forwarders and pass all traffic to servers sitting
in VLAN's behind the F5 (static route point through the core network to the
F5's and DG for the servers). We have had consistent problems for 3 years
with connections being reset due to the TCP profiles that come default from
the appliance. Here is an email response I sent to support (which has to be
F5's weakest point) detailing how we "the customer" fixed a problem we have
been dealing with for quite some time:

 

To support after they suggested the fix we discovered:

 

This is exactly the response I was looking for. We ran across this as well
and enabled it on our core devices, about a week, possibly 2, ago (decided
to see how long it would take for support to get back to us) ;-). 

 

Back in the v4 days of F5, turning on IP forwarding did the exact same
thing, basically turned the system into a stateless IP forwarder. Now, if we
were using F5 for packet inspection and firewall level functions this would
be an issue, but because we have it inline (route hop for servers load
balanced, and a static route on our core network to get to servers that are
load balanced) having the unit perform resets is highly problematic.
Especially when it comes to connection pooling, and this is what I have
forever tried to get across to F5 when it comes to load balancing against
J2EE/JEE based applications (regardless of custom code). 

 

In a J2EE/JEE (or any app that uses connection pooling to a database),
unless the container (JBoss, Weblogic etc) can send interval connection
keepalives to each connection in the pool (typically each container starts
up with a min/max amount of connections), the F5 will automatically shut
them down based on the default tcp profile (300 seconds idle, then a reset
packet). Bumping the idle timeout to an hour simply prolongs the inevitable
and we find the unit reaping connections, that it should care less about,
after it has reached the maximum idle timeout. Additionally an idle timeout
set too high will impact the appliance and actually cause it to fail in a
matter of days.

 

Weblogic handles this well, however, not everyone can afford the cost
associated with Weblogic's pricing model, so we start seeing tomcat and
jboss becoming more and more popular. The problem is, the open source
community has yet to provide a connection pooling health check monitor like
Weblogic or Websphere (at least not that I have seen so far). Until that
becomes available we are stuck having to figure out innovative means of
ensuring the F5 doesn't break the living crap  out of our systems. Second to
that, it doesn't matter if we are using JAVA based applications, .NET or
some other derivative, if it supports connection pooling, and has no
capability of perform interval keepalive health checks in the connection
pool, without loose init/close, we would be in a sorry state.

 

The loose connections (init and close) are what we have been looking for,
for almost 3 YEARS. Can you believe that? I know you don't know me from
Adam, but you are the first support engineer to ever go as deep into a
request and actually provide us with a solution that works. ;-)

 

As you can see there are a ton of amazing features provided by F5, and many
with default settings that if not understood, or understood well, can really
wreak havoc on enterprise systems.

 

I really like F5 products, but I can see why folks get frustrated with them.

 

-Wes

 

 

On Wed, Jul 8, 2009 at 7:44 AM, Kenneth Salchow <[email protected]> wrote:

Julian,

 

I won’t pull a sales pitch—I’m actually interested in knowing if you could
share what those alternate setups are?  Like what’s the business case,
requirements, etc.?  Anyone who has read any of my posts over the last
decade knows that I’m a strong supporter of understanding the problem before
recommending a solution—and no single product can ever be the perfect
solution for every problem.  So, I’m constantly curious as to the real-world
problems that prompt people to say “hey, maybe there’s a better way or
alternative to the way we’ve been doing that”.

 

For one—it just makes me more knowledgeable.   At the same time, if there is
something that we are completely falling short on, this is a good way to
understand what that is and why.

 

Thanks!!

 


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


D 651.423.1133

M 612.868.12588

P 206.272.5555

F 206.272.5555

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

 

 

From: [email protected] [mailto:[email protected]] On Behalf Of
[email protected]
Sent: Wednesday, July 08, 2009 7:55 AM 


To: Load Balancing Mailing List

Subject: Re: [load balancing] Zeus ZXTM?

 

I tested Zeus ZXTM and it worked well for me, and it can do pretty much
everything an F5 LTM will do and had something that had basic iRules
functionality. 



But a new F5 BIG-IP 1600 is under $20k, so ZXTM won't really save you much
money (based on the last time I got prices out of them).

Regards,
Vince





"Julian Grunnell" <[email protected]> 
Sent by: [email protected]

07/08/2009 06:37 AM


Please respond to
Load Balancing Mailing List <[email protected]>


To

"Load Balancing Mailing List" <[email protected]>


cc

	

Subject

[load balancing] Zeus ZXTM?

 

		




Hi – has anyone any experience, good or bad of using Zeus ZXTMor  ZXTM LB?
We almost exclusively use F5 LTM’s at present and have been asked to look at
Zeus as an alternative for some setups.

Thanks – Julian.

Julian Grunnell
UNIX Systems Administrator (Leeds)
Webfusion

Tel: 0208 587 7212
Mob: 07803 649593
Web:  <http://www.webfusion.co.uk/>
http://www.webfusion.co.uk/_______________________________________________
lb-l mailing list
[email protected]
 <http://vegan.net/mailman/listinfo/lb-l>
http://vegan.net/mailman/listinfo/lb-l
Searchable Archive:  <http://vegan.net/lb/archive>
http://vegan.net/lb/archive
 <http://lbdigest.com/> http://lbdigest.comLoad Balancing Digest
 <http://lbwiki.com/> http://lbwiki.comLoad Balancing Wiki


-----------------------------------------------------------------------
This communication, including any attached documentation, is intended only
for the person or entity to which it is addressed, and may contain
confidential, personal, and/or privileged information. Any unauthorized
disclosure, copying, or taking action on the contents is strictly
prohibited. If you have received this message in error, please contact us
immediately so we may correct our records. Please then delete or destroy the
original transmission and any subsequent reply. Thank you.

La présente communication, y compris toute pièce qui y a été jointe, est
destinée uniquement à la personne ou à l’entité à laquelle elle a été
adressée, et contient des renseignements à caractère confidentiel et
personnel. Toute diffusion ou reproduction non autorisée ou toute
intervention entreprise relativement à son contenu est strictement
interdite. Si vous avez reçu ce message par erreur, veuillez nous le
signaler immédiatement afin que nous puissions effectuer la correction à nos
dossiers. Veuillez par la suite supprimer ou détruire le contenu de la
transmission originale ainsi que toute réponse ultérieure. Merci.
-----------------------------------------------------------------------


_______________________________________________
lb-l mailing list
[email protected]
http://vegan.net/mailman/listinfo/lb-l
Searchable Archive: http://vegan.net/lb/archive
http://lbdigest.com <http://lbdigest.com/>  Load Balancing Digest
http://lbwiki.com <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
smime.p7s (application/x-pkcs7-signature, 3 KB) - not displayed