There are some articles on the macromedia site, and there are a few lists on Cf-Talk although mostly are fairly low traffic. I have spent a lot of time over the past 6 months or so working up a decent configuration, and I'm still not done, although I haven't been focusing on it that much until recently.
The problem I have with the configuration that macromedia has is that it's done by proxying with apache servers, and you basically lose the client ip by the time you get to the CF servers. This means that you can't have debugging on your CF servers (or that everyone will see debugging). This doesn't work in our environment, but might be a good fit for you. It also throws out any kind of processing you might want to do with the client IP's. (Like blocking certain ips dynamically, etc). I am looking for a good support group for configuring these servers as well, and would appreciate any leads. Russ > -----Original Message----- > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf > Of John Beynon > Sent: Saturday, January 21, 2006 6:31 AM > To: [email protected] > Subject: Re: [CFCDev] Load balancing and clustering with enterprise > > sadly, once you start talking enterprise stuff there's a lack of > decent documentation since it's a good earner for consultants :) > > I have had much success with mulitple servers load balanced with > windows load balancing and two cf instances on each box...if money > isn't a problem then you go down the hardware route (like > macromedia.com) and have a number of forward facing webservers talking > to remote load balanced CF servers via a hardware load balancer. > > Once you start thinknig about sites with their own JVMs then server > resources are premium - eg I run each instance with 1gb max memory on > 4gb ram servers - I could stretch to a third instance if required but > prefer to leave 2gb memory for the OS. It all depends on the > requirements really and security each requires on the server.... > > best thing is just ask here if you have any more questions, > > john. > > On 1/21/06, Kirk Brogdon <[EMAIL PROTECTED]> wrote: > > Could someone suggest a good source on load balancing, server clustering > and > > using MX 7 enterprise (books, lists, classes)? > > > > We currently run MX 6.1 (server edition) on 2 production web servers > (not > > clustered - one is for Internet sites and the other is for Intranet > sites) > > and also on a development box. From what I have had read about > enterprise, > > I could install it on a completly seperate server and create multiple > > instances for each site so that I could isolate my sites to run in their > own > > jvm. I have a lot of questions and I'm not sure where the best place is > to > > ask them. I have downloaded the administration documentation from > > Macrodobie. > > > > I know this isn't a cfc question but many of you are more familiar with > what > > lists are out there. > > > > Thanks for anything, > > > > Kirk > > > > > > -----Original Message----- > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > > Behalf Of Barney Boisvert > > Sent: Friday, January 20, 2006 4:29 PM > > To: [email protected] > > Subject: Re: [CFCDev] Expense of Operations? > > > > > > A method call is more expensive than a variable reference, but the > > encapsulation the method provides is worth it in almost all scenarios. > > It's very unlikely that this would grow to a bottleneck since the > > more such operations you have, the more other stuff you're doing in > > the request (most likely). Only careful load testing will indicate if > > there's a problem, but I'd say you're safe to assume there won't be. > > > > A few request scoped objects is all you'll need in most cases. Most > > of your objects (including all those you listed in your email) should > > be instantiated into the application and/or session scopes for > > peristance across requests. That'll speed things up because of > > massively reduced instantiation overhead, and let you do stuff like > > caching. > > > > The number of lines of code in an object will affect instantiation of > > the CFC's FIRST instance (when it's read from disk), but not after > > that. The number of methods will affect instantiation of all > > instances (but exceedingly little). Again, until careful load testing > > has illustrated an explicit issue, don't even consider it to be a > > potential issue. > > > > The first and last items are saving you a couple ms tops, per request, > > but the middle item could well be saving you hundreds of ms per > > request. > > > > cheers, > > barneyb > > > > On 1/20/06, Peter Bell <[EMAIL PROTECTED]> wrote: > > > > > > Hello All, > > > > > > Does anyone have any kind of information on the relative expense > > (processing > > > wise) of different operations? For example, assuming a user object has > > been > > > instantiated, how much more processing does it take to > getUserFirstName() > > > (assuming it is just a "return THIS.FirstName") versus just using dot > > > notation against the THIS scoped variable for the object? > > > > > > If you have a few hundred such operations as page of processing a page > > > request, would the choice have any substantive affect on application > > > performance? > > > > > > I know there are guides on CF best practices for performance (avoid > > > evaluate, use list compare instead of multiple string comparisons, > etc.) > > but > > > I can't find a good guide to the relative cost of OO operations in CF > that > > > would help to ensure a design would have a decent shot at being > > performant. > > > > > > Also, while I know it is extremely broad, in general what range of > number > > of > > > request scoped objects mught you instantiate in a common page request > for > > a > > > "well architected app"? I'm assuming that it would be common for a > > > sophisticated page request to instantiate 10-50 objects between the > > > controller, facades, domain objects, compositional objects and the > DAL. Is > > > that a fair heuristic? > > > > > > Finally, does the physical size of an object (the number of lines of > code > > > across all of the methods) have any substantive effect on > initialization > > > performance (does the entire object get loaded into memory or are the > > > methods loaded into memory one at a time? For example, if you have an > > object > > > with 20 methods each with 500 lines of code where any given page > request > > may > > > call only one or two of the methods, would you consider abstracting > some > > of > > > the methods into separate objects to cut down on object instantion > time? > > > > > > I want to clarify, I'm not trying to shave page processing from 45 to > > 18ms, > > > I'm just trying to make sure that pages with minimal db and file > system > > > access don't end up taking 300-400ms. > > > > > > Any thoughts much appreciated! > > > > > > Best Wishes, > > > Peter > > > > -- > > Barney Boisvert > > [EMAIL PROTECTED] > > 360.319.6145 > > http://www.barneyb.com/ > > > > Got Gmail? I have 100 invites. > > > > > > ---------------------------------------------------------- > > You are subscribed to cfcdev. To unsubscribe, send an email to > > [email protected] with the words 'unsubscribe cfcdev' as the subject of > the > > email. > > > > CFCDev is run by CFCZone (www.cfczone.org) and supported by CFXHosting > > (www.cfxhosting.com). > > > > An archive of the CFCDev list is available at > > www.mail-archive.com/[email protected] > > > > > > > > > > > > > > ---------------------------------------------------------- > > You are subscribed to cfcdev. To unsubscribe, send an email to > [email protected] with the words 'unsubscribe cfcdev' as the subject of > the email. > > > > CFCDev is run by CFCZone (www.cfczone.org) and supported by CFXHosting > (www.cfxhosting.com). > > > > An archive of the CFCDev list is available at www.mail- > archive.com/[email protected] > > > > > > > > > -- > <a > href="http://spreadfirefox.com/community/?q=affiliates&id=734&t=1"; > >Get > Firefox!</a> > > > ---------------------------------------------------------- > You are subscribed to cfcdev. To unsubscribe, send an email to > [email protected] with the words 'unsubscribe cfcdev' as the subject of > the email. > > CFCDev is run by CFCZone (www.cfczone.org) and supported by CFXHosting > (www.cfxhosting.com). > > An archive of the CFCDev list is available at www.mail- > archive.com/[email protected] ---------------------------------------------------------- You are subscribed to cfcdev. To unsubscribe, send an email to [email protected] with the words 'unsubscribe cfcdev' as the subject of the email. CFCDev is run by CFCZone (www.cfczone.org) and supported by CFXHosting (www.cfxhosting.com). An archive of the CFCDev list is available at www.mail-archive.com/[email protected]
