From: Noel Bourke (cros13@gmail.com)
Date: Mon Jun 11 2007 - 17:03:52 ART
I'm running Xgl on the laptop.
I can run 24 7200 images no problem,
I just cant really do much else if i do.
CPU utilisation on linux is really not an
indicator of the responsiveness of your system.
What you need to check is the load average.
The load average is based on the number of
non-timer interrupts divided by the number of
minutes over which the average is taken, divided
by 1000/1024. You can see the load average if
you issue the uptime command, or run top or htop.
The three numbers are the load averages over
the past one, five and 15 minutes respectively.
You should compile dynamips yourself for best
stability and performance. Compile with the correct
options for your cpu and threading.
Dynamips can support thing like SSE and for maximum
compatibility these optimisations are not normally
included in most binaries that are distributed.
Good Luck,
Noel
On 6/11/07, kelly@cliffhanger.com <kelly@cliffhanger.com> wrote:
>
> Noel,
>
> Just curious. When you're running those 24
> dynamips routers on the linux box ...do you have
> any 'X' installed or running at the time? Or is
> it soley just a CLI box?
>
> I have a 2.16GHz Core 2 Duo MacBookPro (OSX 10.4.x)
> with 2GB of RAM. When I run 10 to 12 routers,
> it's holding to about 70% CPU utilization.
>
> I'm wondering -- if I dual booted it with a
> minimal installation of a 64bit Linux, without any
> 'X', if I could get close to the performance you
> are reporting. I realize you have 2GB more or
> RAM, so I'm sure it won't be exactly the same
> ...but maybe a good deal better than I get now.
>
>
> --
> ___
> /\ \
> / \ \
> / \ \
> / /\ \ \
> / / \ \ \
> / / / \ \ \
> / / /___\__\ \
> / / /___________\
> \/_______________/
>
> Impossible Triangle
> M. C. Escher
>
> Big riffs, massive grooves, and expansive improvisations
>
>
>
>
> Quoting Niche <jackyliu419@gmail.com>:
> /bow.. thank you very much for the explanation~
>
> Cheers~
> Jacky
>
> On 6/11/07, Noel Bourke <cros13@gmail.com> wrote:
> >Depends on what xeon models you are talking about, they can have
> varied
> >architectures.
> >The 3/5/7000 series xeons are based on the same architecture as
> the core 2
> >with more cache.
> >The 3.2Ghz xeon noncona's i use on my fixed dynamips system at
> home are
> >essentially 64 bit prescott
> >P4s with more cache (I havent been home since november, so no
> point in
> >upgrading)
> >I use a 2.33 core 2 in my laptop and run dynamips on that.
> >
> >Really there are so many generations and varients of xeon that
> there is
> >little i can say that would be
> >applicable to your question.
> >
> >There are some generalisations i can make.
> >Xeons are always based on a consumer architecture (e.g. Xeon
> Noncona = P4
> >Prescott,
> >Xeon Paxville = PD Smithfield, Xeon Woodcrest = Core 2 Duo
> (Conroe) ) and
> >they inherit
> >mostly the performance and characteristics of the consumer arch
> on which
> >they are based.
> >Modern Xeons are basically dual/multi processor intact (on the
> consumer
> >versions this is forced disabled)
> >with extra cache (cache is expensive) and maybe a different
> pinout to help
> >(force) you to buy a
> >new motherboard with one of the expensive intel workstation
> chipsets.
> >http://en.wikipedia.org/wiki/Xeon should explain the
> >differences best.
> >
> >As to the choice.... your quad core consumer chip is really two
> core 2 duos
> >packaged together.
> >It will be cheaper, use less power, be more reliable, not force
> you to buy
> >a
> >chipset/motherboard combo
> >you dont want or need for 5 times the price of a consumer board
> with the
> >same features as far as your use goes,
> >run at a higher fsb speed than any but the quad core xeon due to
> it being a
> >kentsfield chip giving a slight bump in performance.
> >
> >In short there are plenty of reasons to go quad-core instead of
> xeon, and
> >unless you are building a dual
> >quad-core xeon rig (if you are tell me who pays you that much so
> i can
> >replace you ;) ) there is really
> >no reason to go xeon unless you like the logo and big bills.
> >
> >Regards,
> > Noel
> >
> >
> > On 6/11/07, Niche <jackyliu419@gmail.com> wrote:
> >> Hi Noel,
> >>
> >> Just asking,
> >>
> >> 1. Intel Quad Core CPU
> >> 2. Dual-Core Xeon x 2
> >>
> >> I know it's hard to do a direct comparison but let pick the
> closest
> >> match for clock rate of the CPU for both setup. Which one will
> going
> >> to provide the best performance? I was told by one of my friend
> that
> >> Setup 2 will win.
> >>
> >> Cheers~
> >> Jacky
> >>
> >> > Quoting Noel Bourke <cros13@gmail.com>:
> >> > Core 2, no question. I'm simplifying a bit here
> >> > with my explanation of why, and glossing over
> >> > quite a few issues.
> >> >
> >> > Core 2 is a "four wide" architecture, meaning that
> >> > it can issue four instructions per clock cycle.
> >> > All of AMD's products are "3 wide". Predictability
> >> > of instructions is crucial here to gain maximum
> >> > benefit.
> >> >
> >> > Also due to the that emulation generates quite
> >> > predictable instructions, dynamips benefits from a
> >> > longer pipeline. This gives the core 2's 14-stage
> >> > an advantage over the AMD 12-stage. Its also the
> >> > reason why the 31-stage p4 (netburst) did so well
> >> > with binary patching (e.g. VMware) and
> >> > architecture emulation (e.g. PearPC).
> >> >
> >> > The Core2's dynamically allocated shared cache
> >> > also helps. More cache can be used by a core
> >> > running a dynamips process which is under greater
> >> > load (e.g. emulating say a hub router, or a router
> >> > running more protocols than the others). AMD has
> >> > the hypertransport bus and onboard memory
> >> > controller which is not as decisive an advantage
> >> > as the core 2 cache.
> >> >
> >> > A dynamips can also be aware of data loaded to the
> >> > cache by a process running on the other core which
> >> > with a static image file like the ios image You
> >> > are using with dynamips can reduce the amount of
> >> > times the process has to access the drastically
> >> > slower main memory.
> >> >
> >> > So to simplify a bit further, say your dynamips
> >> > process on one core is running though the section
> >> > of the ios image to send an ospf hello. 1.3
> >> > seconds later another dynamips process running on
> >> > the other core has to send a hello too, your AMD
> >> > has to take a comparatively long commute to the
> >> > memory controller (yes even though the memory
> >> > controller is onboard its still further then the
> >> > cache), and then an epic transcontinental journey
> >> > to main memory. For all it cares your other cpu
> >> > cores cache could be in siberia instead of right
> >> > beside it on the die, its not even aware of its
> >> > existence let alone contents. The AMD will
> >> > probably have half or less cache then the core 2
> >> > regardless.
> >> >
> >> > OK enough on the CPU. Core 2 it is then.
> >> >
> >> > RAM....uhuh....ok....go for the fastest you can
> >> > get (PC5200+). no manufacturer or retailer i know
> >> > of tells you any more specifics, such as timings
> >> > so ignore the rest.
> >> >
> >> > As to amount of RAM....errr i'm going to say this
> >> > later but any 32 bit OS is limited to 4GB of
> >> > memory (excluding PAE but that isent relevant as 4
> >> > GB 'aint the major barrier. On windows memory is
> >> > handled pretty awfully. what happen when you have4
> >> > gigs of memory is say for example on 32bit XP,
> >> > each process is essentially limited to 1.8 gb.
> >> >
> >> > Last point, don't run dynamips on windows. Its
> >> > running not quite natively on an architecture
> >> > which bears little resemblence to the robust unix
> >> > box it was designed for, which happens to have
> >> > awful memory management, a bolted on network
> >> > stack, pitiful management of multiple processors
> >> > and multithreading that grinds to a halt when
> >> > under load. thats not microsoft-bashing, windows
> >> > is severely structurally flawed due to the bolting
> >> > on of not just features but entire concepts (like
> >> > multi-user and networking), both apply to
> >> > dynamips.
> >> >
> >> > Linux will run daynamips with the same configs
> >> > 10-20% faster on any processor. and it won't
> >> > become unusable at 100% cpu usage because
> >> > processes are pretty much forced to play nice.
> >> >
> >> > If you are voluntarily using dynamips on vista i'd
> >> > call the nice doctors in the white coats or you
> >> > could get a job far enough away from computers
> >> > that you cant hurt them anymore.
> >> >
> >> > Do
> >> > - use any 64bit distro of linux,
> >> > - do download the source code from the dynamips
> site,
> >> > - do compile using
> >> > gcc 3.2+ using cpu specific optimisation
> >> > and the -O3 option for threading,
> >> > - do set your idle-pc correctly.
> >> >
> >> > I emulate up to 24 7200s simultaneously here on my
> >> > laptop (2.33Ghz Core 2, 4GB Ram).
> >> >
> >> > Apologies for the quasi-religous ferver.
> >> >
> >> > Regards,
> >> > Noel Bourke
> >> >
> >> > On 6/9/07, Con Spathas <con@spathas.net> wrote:
> >> > >
> >> > > Check out http://7200emu.hacki.at/
> >> > >
> >> > > This question has been asked many times there!
> >> > >
> >> > > Cheers...
> >> > >
> >> > > -----Original Message-----
> >> > > From: nobody@groupstudy.com [mailto:
> nobody@groupstudy.com] On
> >Behalf Of
> >> > > lalit gupta
> >> > > Sent: Saturday, 9 June 2007 08:19
> >> > > To: Cisco certification
> >> > > Subject: Maximum number of instances on
> dynamips/dynagen.u
> >> > > I am planning to buy a laptop for running maximum
> number of
> >instances on
> >> > > dynamips/dynagen.
> >> > >
> >> > > Which configuration will you geniuses recommend.
> >> > >
> >> > > 1) AMD 64-bit process 2GHz
> >> > > 2 GB RAM
> >> > > or
> >> > > 2) Core 2 Duo processor 2 GHz
> >> > > 2 GB RAM
> >> > >
> >> > > Also will it help if I increase the RAM further
> more.
> >> > >
> >> > > Please if you can given any more information which
> will be
> >useful please
> >> > > don't hesitate to reply.....
> >> > >
> >> > > Regards
> >> > > lalit
> >> > >
> >> > >
>
> >_______________________________________________________________________
> >> > > Subscription information may be found at:
> >> > > http://www.groupstudy.com/list/CCIELab.html
> >> > >
> >> > >
>
> >_______________________________________________________________________
> >> > > Subscription information may be found at:
> >> > > http://www.groupstudy.com/list/CCIELab.html
> >> >
> >> >
>
> >_______________________________________________________________________
> >> > Subscription information may be found at:
> >> > http://www.groupstudy.com/list/CCIELab.html
> >> >
> >> >
>
> >_______________________________________________________________________
> >> > Subscription information may be found at:
> >> > http://www.groupstudy.com/list/CCIELab.html
>
>
> _______________________________________________________________________
> Subscription information may be found at:
> http://www.groupstudy.com/list/CCIELab.html
This archive was generated by hypermail 2.1.4 : Sun Jul 01 2007 - 17:24:48 ART