«

Hi,

I have to load test a switch(content switch - CSS 11050) for work. I
am currently using a software called http_load. It's producing
requests to a web site that passes through the switch. That way I got
the cpuload up to about 80%. But I can't get it higher!

Does anyone had to do this before?
Any idea on how to increase the cpuload of a switch to 100% is
welcomed!!

Thank you so much...
Genevieve

zanzan wrote:

> Hi,
>
> I have to load test a switch(content switch - CSS 11050) for work. I
> am currently using a software called http_load. It's producing
> requests to a web site that passes through the switch. That way I got
> the cpuload up to about 80%. But I can't get it higher!
>
> Does anyone had to do this before?
> Any idea on how to increase the cpuload of a switch to 100% is
> welcomed!!
>
> Thank you so much...
> Genevieve

How much bandwidth do you have left in the pipe? It's always possible that
the CSS 11050 was designed such that the pipe will be saturated before the
CPU is overrun (And I'd have to count that as an eminently reasonable
design decision). Otherwise, I'd look at adding a couple of machines to
your testbed, and loading http_load or similar onto them.

Jeff

I am not sure the intention of this request, but.....
There are million of ways to make your CSS CPU reaches 100%.

Ethical way is to use numerous small udp packets.....
Evil way is to use DOS tools.


"zanzan" <> ???
news: om ???...
> Hi,
>
> I have to load test a switch(content switch - CSS 11050) for work. I
> am currently using a software called http_load. It's producing
> requests to a web site that passes through the switch. That way I got
> the cpuload up to about 80%. But I can't get it higher!
>
> Does anyone had to do this before?
> Any idea on how to increase the cpuload of a switch to 100% is
> welcomed!!
>
> Thank you so much...
> Genevieve


--
~ Samba, more than a low cost File and Printer server ~

-- Let us OpenSource --


-----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
-----== Over 100,000 Newsgroups - 19 Different Servers! =-----

Thanks,
But...
What are upd packets and how can I do that?


"Dede" <> wrote in message news:<>...
> I am not sure the intention of this request, but.....
> There are million of ways to make your CSS CPU reaches 100%.
>
> Ethical way is to use numerous small udp packets.....
> Evil way is to use DOS tools.
>
>
> "zanzan" <> ???
> news: om ???...
> > Hi,
> >
> > I have to load test a switch(content switch - CSS 11050) for work. I
> > am currently using a software called http_load. It's producing
> > requests to a web site that passes through the switch. That way I got
> > the cpuload up to about 80%. But I can't get it higher!
> >
> > Does anyone had to do this before?
> > Any idea on how to increase the cpuload of a switch to 100% is
> > welcomed!!
> >
> > Thank you so much...
> > Genevieve
>
>
> --
> ~ Samba, more than a low cost File and Printer server ~
>
> -- Let us OpenSource --
>
>
> -----= Posted via Newsfeeds.Com, Uncensored Usenet News =-----
> http://www.newsfeeds.com - The #1 Newsgroup Service in the World!
> -----== Over 100,000 Newsgroups - 19 Different Servers! =-----

Thanks..

I had 2 computers using http_load when I first started. I added one
more but it didn't change anything....I'm still using the 3 computers
cause we never know!!

How can I check about the bandwidth thing?

Jeff Pratt <> wrote in message news:<8elZc.42405$A8.37215@edtnps89>...
> zanzan wrote:
>
> > Hi,
> >
> > I have to load test a switch(content switch - CSS 11050) for work. I
> > am currently using a software called http_load. It's producing
> > requests to a web site that passes through the switch. That way I got
> > the cpuload up to about 80%. But I can't get it higher!
> >
> > Does anyone had to do this before?
> > Any idea on how to increase the cpuload of a switch to 100% is
> > welcomed!!
> >
> > Thank you so much...
> > Genevieve
>
> How much bandwidth do you have left in the pipe? It's always possible that
> the CSS 11050 was designed such that the pipe will be saturated before the
> CPU is overrun (And I'd have to count that as an eminently reasonable
> design decision). Otherwise, I'd look at adding a couple of machines to
> your testbed, and loading http_load or similar onto them.
>
> Jeff

(zanzan) wrote in message news:<. com>...
> Thanks,
> But...
> What are upd packets and how can I do that?
>
>
> "Dede" <> wrote in message news:<>...
> > I am not sure the intention of this request, but.....
> > There are million of ways to make your CSS CPU reaches 100%.
> >
> > Ethical way is to use numerous small udp packets.....
> > Evil way is to use DOS tools.
> >
> >
> > "zanzan" <> ???
> > news: om ???...
> > > Hi,
> > >
> > > I have to load test a switch(content switch - CSS 11050) for work. I
> > > am currently using a software called http_load. It's producing
> > > requests to a web site that passes through the switch. That way I got
> > > the cpuload up to about 80%. But I can't get it higher!
> > >
> > > Does anyone had to do this before?
> > > Any idea on how to increase the cpuload of a switch to 100% is
> > > welcomed!!

Hello,

You need to find out where the limit to the number of transactions
(let's call them) is occurring. It could be the clients, the servers,
the network, the CSS box.

Add more clients and servers, make sure the network is not
getting in the way, errors, bandwidth limits.

It is possible that the CSS thingy is designed to always leave
some CPU for housekeeping and it may not be possible to saturate it
with http requests.

The reference to UDP packets is not likely to be appropriate for
your testing. It is an alternative way of saturating the box's
CPU with what is effectively rubbish traffic that has no meaning
other than to demonstrate the devices resistance to certain
denial of service attacks. My guess would be that you are
intending to test the device against http traffic.

http_load has a limit of 1000 requests per second, I would like to
raise that number, but that's the limit...and I execute http_load on 3
computers.
I don't think that the CSS keeps some cpu for housekeeping...because
I've seen it up to 94% on a special occasion.
Also, the web page I use for http_load's requests is very small, I
have only one line saying "this is a small page..."...I found out that
with a bigger page the switch takes too much time to treat it. And the
cpu load wasn't very high. Now with smaller pages, my cpu load is
around 80% like I was saying...

The reason I have to do this project is because we are receiving a
newer switch and we want to see how good this new switch is compared
to the one we have now.
So when we'll get the new one, I'll do the same tests and hopefully
the cpu load won't be very high! Do you understand? It's true that we
want to test it against http traffic...we thought about this project
because we had a lot of users at one time on the web site and the cpu
got up to 92% and we freaked out! We would also like to find out what
would happen if it got to 100%...



(AnyBody43) wrote in message news:<. com>...
> (zanzan) wrote in message news:<. com>...
> > Thanks,
> > But...
> > What are upd packets and how can I do that?
> >
> >
> > "Dede" <> wrote in message news:<>...
> > > I am not sure the intention of this request, but.....
> > > There are million of ways to make your CSS CPU reaches 100%.
> > >
> > > Ethical way is to use numerous small udp packets.....
> > > Evil way is to use DOS tools.
> > >
> > >
> > > "zanzan" <> ???
> > > news: om ???...
> > > > Hi,
> > > >
> > > > I have to load test a switch(content switch - CSS 11050) for work. I
> > > > am currently using a software called http_load. It's producing
> > > > requests to a web site that passes through the switch. That way I got
> > > > the cpuload up to about 80%. But I can't get it higher!
> > > >
> > > > Does anyone had to do this before?
> > > > Any idea on how to increase the cpuload of a switch to 100% is
> > > > welcomed!!
>
> Hello,
>
> You need to find out where the limit to the number of transactions
> (let's call them) is occurring. It could be the clients, the servers,
> the network, the CSS box.
>
> Add more clients and servers, make sure the network is not
> getting in the way, errors, bandwidth limits.
>
> It is possible that the CSS thingy is designed to always leave
> some CPU for housekeeping and it may not be possible to saturate it
> with http requests.
>
> The reference to UDP packets is not likely to be appropriate for
> your testing. It is an alternative way of saturating the box's
> CPU with what is effectively rubbish traffic that has no meaning
> other than to demonstrate the devices resistance to certain
> denial of service attacks. My guess would be that you are
> intending to test the device against http traffic.

(zanzan) wrote in message news:<. com>...
> http_load has a limit of 1000 requests per second, I would like to
> raise that number, but that's the limit...and I execute http_load on 3
> computers.
> I don't think that the CSS keeps some cpu for housekeeping...because
> I've seen it up to 94% on a special occasion.
> Also, the web page I use for http_load's requests is very small, I
> have only one line saying "this is a small page..."...I found out that
> with a bigger page the switch takes too much time to treat it. And the
> cpu load wasn't very high. Now with smaller pages, my cpu load is
> around 80% like I was saying...
>
> The reason I have to do this project is because we are receiving a
> newer switch and we want to see how good this new switch is compared
> to the one we have now.
> So when we'll get the new one, I'll do the same tests and hopefully
> the cpu load won't be very high! Do you understand? It's true that we
> want to test it against http traffic...we thought about this project
> because we had a lot of users at one time on the web site and the cpu
> got up to 92% and we freaked out! We would also like to find out what
> would happen if it got to 100%...

I don't know (or care) what exactly a CSS or http_load are.

You alredy seem to have the data point that you need.

http_load configured like A
Request rate B
CSS CPU C

Put in the new box and you will have

http_load configured like A
Request rate X
New thing CPU Y


Hopefully X > B

and

Y/X < C/B


As to what happens as CPU approaches 100%, well who knows?
Some circumstances like that seem to have a gracefull failure
whereby the service provided just reaches its capacity however
as I am sure you have noticed sometimes computers fail less gracefully
when resources are scarce.

On 3 Sep 2004 09:55:55 -0700, (zanzan) wrote:

>http_load has a limit of 1000 requests per second, I would like to
>raise that number, but that's the limit...and I execute http_load on 3
>computers.

This is a good thing. Otherwise, someone can Denial-of-Service attack
you. You might also setup a script to constantly query snmp tables,
or setup insane ACLs, etc.

I will try ACLs if I can..

Any more ideas anyone?

Thanks
Genevieve

wrote in message news:<>. ..
> On 3 Sep 2004 09:55:55 -0700, (zanzan) wrote:
>
> >http_load has a limit of 1000 requests per second, I would like to
> >raise that number, but that's the limit...and I execute http_load on 3
> >computers.
>
> This is a good thing. Otherwise, someone can Denial-of-Service attack
> you. You might also setup a script to constantly query snmp tables,
> or setup insane ACLs, etc.