Renuan:

Thanks for taking the time.

Jack

--- In [email protected], "kaibabsowats" <[EMAIL PROTECTED]> 
wrote:
> The number was a "at least" number, so it was definetely low.
> 
> I'll take a stab and trying to explain all the depends, but I 
might be
> off on some of these.
> 
> Performance hit - Flex compiling the MXML into SWF files, there is
> caching mechanisms built in, dont know the extent how well the 
caching
> works.  This process is usaully CPU then Memory intensive.  This is
> done when in the JVM so your bottleneck here is the JVM used (IBM,
> Sun, BEA Jrockit, etc... ).
> 
> Backend Data Tier ( WebServices/Remoting etc... ) - This depends on
> what you end up using, could be Java, .Net, or ColdFusion (even 
PHP or
> Perl).  Assuming you run any of these on the same machine you'll 
have
> performance in CPU or Memory, where these bottlenecks follow a 
typical
> application bottleneck.  These service's performance is not 
directly
> related to Flex, unless you are using Flex's sandbox policy then 
there
> would be some overhead placed on the JVM again to handle each 
request.
> 
> As for benchmarking there are software programs out there that can
> replay http requests and simulate lots of people hitting your 
servers.
>   But its still hard to get good numbers unless you simulate your 
real
> deployment environment as close as possible.  
> 
> One thing going for Flex is the idea that instead of having alot of
> small requests hitting your server over a period of time, which 
eats
> up resources, you have one medium/big hit at first with little
> interaction with the server after (of course depending on if the
> application is getting real-time data or not).  You have a 
situation
> where they hit it less times over a period of time thus off setting
> the time server resource you give up for running Flex.
> 
> The disclaimer is that it just depends.  Depends on size of the 
SWF's,
> complexity of MXML Flex compiling, data being transfered, 
interaction
> of user (does the app have real time data ), etc...  
> 
> In my case I am doing online video training and my bandwidth is 
going
> to be a bigger issue then Flex performance for the number of 
projected
> users.
> 
> Its really hard to say what a specific server can do without 
knowing
> the extent of how its going to be used and the application being
> built.
> 
> Renaun
> 
> --- In [email protected], "jwc_wensan" <[EMAIL PROTECTED]>
> wrote:
> > Renaun:
> > 
> > I appreciate your input.
> > 
> > That number seems low.  What do you see is the bottleneck?  Flex,
> > CPU, RAM, etc.?
> > 
> > Without going live, any suggestions on how to benchmark?
> > 
> > Thanks,
> > 
> > Jack
> > 
> > 
> > --- In [email protected], "kaibabsowats" <[EMAIL PROTECTED]> 
> > wrote:
> > > I won't give you definite answer as it does really depend, but
> give
> > > you a at least number.  
> > > 
> > > With Java the JVM has a 1.5Gb ram limitation (unless using Sun 
> > boxes
> > > or 64 bit correct me if I am wrong).  So I usually try and get
> 2Gb 
> > of
> > > ram minimum.  This allows the JVM use its limit if needed and 
have
> > > stuff left over for normal OS and Database usage.
> > > 
> > > The number I can give is that you can definitely handle a 
couple
> > > hundred simultaneous users on a server like that.  To spec out
> > > anything more you need to benchmark the server with the actual
> > > application.  I had a Dual 3.2Ghz Xeon 2Gb ram handle 1000-2000
> > > simultaneous users for a Java application that was Memory 
> > intensive.
> > > 
> > > It is pretty safe to say if you actually have a couple hundred
> > > simultaneous users you mostly likely have ten of thousand 
actual
> > > users.  Of course there is no magic number but these are given 
as 
> > is
> > > and really should be benchmarked with the actual application.
> > > 
> > > Renaun
> > > 
> > > --- In [email protected], "jwc_wensan" 
<[EMAIL PROTECTED]> 
> > wrote:
> > > > To All:
> > > > 
> > > > Yesterday I made a post regarding trying to determine how 
many 
> > users
> > > > a single server with one CPU could handle.  I realize "it 
> > depends"
> > > > and I did not define any type of application.
> > > > 
> > > > I am trying to get some type of handle on what my costs will 
be
> > > > regarding CPU licenses, servers, etc.
> > > > 
> > > > So let me try again.  Let's take the Flexstore as an example 
> > > > application.  Based on this, can anyone shed some light as 
to 
> > how 
> > > > many users in an hour the server could adequately handle.
> > > > 
> > > > Where will the bottleneck occur: Flex, server itself, 
database, 
> > etc.
> > > > 
> > > > If an hour is not the right way to measure, then please offer
> any
> > > > measurement you feel is correct.
> > > > 
> > > > Server Configuration:
> > > > 
> > > >     Dell Server @ 3.8 GHz with 1 GB RAM
> > > >     JRUN4
> > > >     Win 2000 Server
> > > >     1 Flex license
> > > > 
> > > > As Dave and others recommended, I will put the database on a 
> > separate
> > > > server.
> > > > 
> > > > I am not trying to tie anyone's hands here, just trying to 
get 
> > some
> > > > perspective.
> > > > 
> > > > Thanks in advance,
> > > > 
> > > > Jack




--
Flexcoders Mailing List
FAQ: http://groups.yahoo.com/group/flexcoders/files/flexcodersFAQ.txt
Search Archives: http://www.mail-archive.com/flexcoders%40yahoogroups.com 
Yahoo! Groups Links

<*> To visit your group on the web, go to:
    http://groups.yahoo.com/group/flexcoders/

<*> To unsubscribe from this group, send an email to:
    [EMAIL PROTECTED]

<*> Your use of Yahoo! Groups is subject to:
    http://docs.yahoo.com/info/terms/
 



Reply via email to