Re: [tlug] Re: Talk about fast HTTP

[Edward Middleton <emiddleton@example.com>]

Sach Jobb wrote:
> This conversation seems to have gone a bit off topic. Bringing things back to the original discussion...
>
> > > I wonder if it's worth having a panel discussion after the talk.
> > > It sounds as if there may be several of us who design and build
> > > high-performance HTTP systems who might have useful experience to
> > share. 
>
> A panel could be interesting as well. Not everyone in TLUG works directly with HTTP on a daily basis but from a bit of an outside observation it seems that there are quite a few people that do. Perhaps this panel could be held at a separate venue which could be accommodating to all by providing a facility that would allow for the purchase and consumption of alcoholic beverages.

I don't know how well this sort of discussion is going to work in a 
panel.  Performance tuning is very dependent on the type of application, 
if we don't limit the scope we will end up discussing a lot of things 
with far too little details to be useful.  I guess we could all discuss 
John Femlins' approach.

Another idea might be to have panelists give a short 5-10 minute 
introduction to a particular aspect/application of performance 
optimization and build a panel discussion based on this.

It should also be noted that in real world applications, network latency 
can be far more important then application latency.  A case in point, 
my blog[1] using varnish serves cached files in 10E-4 seconds, but from 
the US it can take 6 seconds to load a cached page.  This is almost 
entirely download and round trip delay.  Obviously CDN's or hosting near 
to you audience can help reduce this as well as correctly setting expire 
headers and combining JavaScript, CSS where appropriate, but it is still 
a factor that can't be ignored.

There are also other issues like reducing user observable latency which 
can be more important then average system latency.

> > This could be interesting if you remove the "HTTP" word. I believe
> > that
> > most companies are looking for performance in networking (I really
> > mean system
> > and software tuning here, not physical limits of the hardware/lines),
> > whatever
> > the TCP protocol is.
>
> Although performance in general is always an interesting topic, I think if we open it up it might be a bit too broad. Is it OK if we just keep it focused on web application performance?

I have been using thrift[2] quite a bit lately and would be interested 
in hearing about general networking optimization.

Edward

1. http://blogs.vortorus.net/
2. http://incubator.apache.org/thrift/