From: Lars Rune Nøstdal
Subject: Re: where is the lisp operating system? (new http server/backend for lisp)
Date:
Message-ID: <1222079439.10480.453.camel@blackbox>
On Mon, 2008-09-22 at 08:46 +0200, Lars Rune Nøstdal wrote:
> On Sat, 2008-09-20 at 21:37 -0700, gavino wrote:
> > where is the lisp operating system?
> >
> > If lisp is more powerful........then why is there not a working lispos
> > with a desktop and broswer?
>
> *rant*
Ok..that was ranty. Anyway, I'm building a new low-level HTTP backend
for SW, but this can be used for any Lisp type HTTP server. Latest test
run:
* 300 000 000 request/response round-trips (dynamically generated pages)
since yesterday; it's stable and doesn't leak memory or resources (FDs
etc.).
* 5000 concurrent sessions or always-connected clients (using HTTP
keep-alive etc.); it scales. This is probably a modest number since
normally each of the 5000 clients doesn't _constantly_ hammer the server
like I do in this test.
* ~5000 concurrent request/response round-trips pr. second with dynamic
content; it's fast.
The 5000 clients _and_ the server is currently running on the _same_
machine (old junk; standard AMD desktop hardware from late 2006). The
clients steal a lot of CPU from the server so I'll probably get higher
numbers when I get hold of a second machine for the clients.
It's based on IOLib (uses Linux epoll stuff) and the source is here:
http://common-lisp.net/~lnostdal/programming/lisp/sw-http/
The reason I'm writing this is because I need real scalability in a
"comet" or "reversed AJAX" scenario where the amount of always-connected
concurrent clients/sessions is way beyond what normal HTTP servers are
designed for.
Note that this thing is still work in progress..
..but I think this will kick some ass. When I move the clients to a
different machine it will probably handle 10000 - 20000 concurrent
always-connected clients/sessions just fine.
The idea is to keep this very low-level, fast and "scalable"; and to not
parse most things unless the user explicitly asks us to etc.
Bah .. this was ranty too .. never mind .. :}
--
Lars Rune Nøstdal || AJAX/Comet GUI type stuff for Common Lisp
http://nostdal.org/ || http://groups.google.com/group/symbolicweb
On Sep 22, 3:30 am, Lars Rune Nøstdal <···········@gmail.com> wrote:
> On Mon, 2008-09-22 at 08:46 +0200, Lars Rune Nøstdal wrote:
> > On Sat, 2008-09-20 at 21:37 -0700, gavino wrote:
> > > where is the lisp operating system?
>
> > > If lisp is more powerful........then why is there not a working lispos
> > > with a desktop and broswer?
>
> > *rant*
>
> Ok..that was ranty. Anyway, I'm building a new low-level HTTP backend
> for SW, but this can be used for any Lisp type HTTP server. Latest test
> run:
>
> * 300 000 000 request/response round-trips (dynamically generated pages)
> since yesterday; it's stable and doesn't leak memory or resources (FDs
> etc.).
>
> * 5000 concurrent sessions or always-connected clients (using HTTP
> keep-alive etc.); it scales. This is probably a modest number since
> normally each of the 5000 clients doesn't _constantly_ hammer the server
> like I do in this test.
>
> * ~5000 concurrent request/response round-trips pr. second with dynamic
> content; it's fast.
>
> The 5000 clients _and_ the server is currently running on the _same_
> machine (old junk; standard AMD desktop hardware from late 2006). The
> clients steal a lot of CPU from the server so I'll probably get higher
> numbers when I get hold of a second machine for the clients.
>
> It's based on IOLib (uses Linux epoll stuff) and the source is here:http://common-lisp.net/~lnostdal/programming/lisp/sw-http/
>
> The reason I'm writing this is because I need real scalability in a
> "comet" or "reversed AJAX" scenario where the amount of always-connected
> concurrent clients/sessions is way beyond what normal HTTP servers are
> designed for.
>
> Note that this thing is still work in progress..
>
> ..but I think this will kick some ass. When I move the clients to a
> different machine it will probably handle 10000 - 20000 concurrent
> always-connected clients/sessions just fine.
>
> The idea is to keep this very low-level, fast and "scalable"; and to not
> parse most things unless the user explicitly asks us to etc.
>
> Bah .. this was ranty too .. never mind .. :}
>
> --
> Lars Rune Nøstdal || AJAX/Comet GUI type stuff for Common Lisphttp://nostdal.org/||http://groups.google.com/group/symbolicweb
WOW
AWESOME!!!
On Sep 22, 5:48 am, gavino <·········@gmail.com> wrote:
> On Sep 22, 3:30 am, Lars Rune Nøstdal <···········@gmail.com> wrote:
>
> > On Mon, 2008-09-22 at 08:46 +0200, Lars Rune Nøstdal wrote:
> > > On Sat, 2008-09-20 at 21:37 -0700, gavino wrote:
> > > > where is the lisp operating system?
>
> > > > If lisp is more powerful........then why is there not a working lispos
> > > > with a desktop and broswer?
>
> > > *rant*
>
> > Ok..that was ranty. Anyway, I'm building a new low-level HTTP backend
> > for SW, but this can be used for any Lisp type HTTP server. Latest test
> > run:
>
> > * 300 000 000 request/response round-trips (dynamically generated pages)
> > since yesterday; it's stable and doesn't leak memory or resources (FDs
> > etc.).
>
> > * 5000 concurrent sessions or always-connected clients (using HTTP
> > keep-alive etc.); it scales. This is probably a modest number since
> > normally each of the 5000 clients doesn't _constantly_ hammer the server
> > like I do in this test.
>
> > * ~5000 concurrent request/response round-trips pr. second with dynamic
> > content; it's fast.
>
> > The 5000 clients _and_ the server is currently running on the _same_
> > machine (old junk; standard AMD desktop hardware from late 2006). The
> > clients steal a lot of CPU from the server so I'll probably get higher
> > numbers when I get hold of a second machine for the clients.
>
> > It's based on IOLib (uses Linux epoll stuff) and the source is here:http://common-lisp.net/~lnostdal/programming/lisp/sw-http/
>
> > The reason I'm writing this is because I need real scalability in a
> > "comet" or "reversed AJAX" scenario where the amount of always-connected
> > concurrent clients/sessions is way beyond what normal HTTP servers are
> > designed for.
>
> > Note that this thing is still work in progress..
>
> > ..but I think this will kick some ass. When I move the clients to a
> > different machine it will probably handle 10000 - 20000 concurrent
> > always-connected clients/sessions just fine.
>
> > The idea is to keep this very low-level, fast and "scalable"; and to not
> > parse most things unless the user explicitly asks us to etc.
>
> > Bah .. this was ranty too .. never mind .. :}
>
> > --
> > Lars Rune Nøstdal || AJAX/Comet GUI type stuff for Common Lisphttp://nostdal.org/||http://groups.google.com/group/symbolicweb
>
> WOW
> AWESOME!!!
Congratulations, Lara! It is fitting that you should have a fanboy
such as gavino.
From: Lars Rune Nøstdal
Subject: Re: where is the lisp operating system? (new http server/backend for lisp)
Date:
Message-ID: <1222116558.10480.529.camel@blackbox>
On Mon, 2008-09-22 at 12:59 -0700, William James wrote:
> On Sep 22, 5:48 am, gavino <·········@gmail.com> wrote:
> > On Sep 22, 3:30 am, Lars Rune Nøstdal <···········@gmail.com> wrote:
> >
> > > On Mon, 2008-09-22 at 08:46 +0200, Lars Rune Nøstdal wrote:
> > > > On Sat, 2008-09-20 at 21:37 -0700, gavino wrote:
> > > > > where is the lisp operating system?
> >
> > > > > If lisp is more powerful........then why is there not a working lispos
> > > > > with a desktop and broswer?
> >
> > > > *rant*
> >
> > > Ok..that was ranty. Anyway, I'm building a new low-level HTTP backend
> > > for SW, but this can be used for any Lisp type HTTP server. Latest test
> > > run:
> >
> > > * 300 000 000 request/response round-trips (dynamically generated pages)
> > > since yesterday; it's stable and doesn't leak memory or resources (FDs
> > > etc.).
> >
> > > * 5000 concurrent sessions or always-connected clients (using HTTP
> > > keep-alive etc.); it scales. This is probably a modest number since
> > > normally each of the 5000 clients doesn't _constantly_ hammer the server
> > > like I do in this test.
> >
> > > * ~5000 concurrent request/response round-trips pr. second with dynamic
> > > content; it's fast.
> >
> > > The 5000 clients _and_ the server is currently running on the _same_
> > > machine (old junk; standard AMD desktop hardware from late 2006). The
> > > clients steal a lot of CPU from the server so I'll probably get higher
> > > numbers when I get hold of a second machine for the clients.
> >
> > > It's based on IOLib (uses Linux epoll stuff) and the source is here:http://common-lisp.net/~lnostdal/programming/lisp/sw-http/
> >
> > > The reason I'm writing this is because I need real scalability in a
> > > "comet" or "reversed AJAX" scenario where the amount of always-connected
> > > concurrent clients/sessions is way beyond what normal HTTP servers are
> > > designed for.
> >
> > > Note that this thing is still work in progress..
> >
> > > ..but I think this will kick some ass. When I move the clients to a
> > > different machine it will probably handle 10000 - 20000 concurrent
> > > always-connected clients/sessions just fine.
> >
> > > The idea is to keep this very low-level, fast and "scalable"; and to not
> > > parse most things unless the user explicitly asks us to etc.
> >
> > > Bah .. this was ranty too .. never mind .. :}
> >
> > > --
> > > Lars Rune Nøstdal || AJAX/Comet GUI type stuff for Common Lisphttp://nostdal.org/||http://groups.google.com/group/symbolicweb
> >
> > WOW
> > AWESOME!!!
>
> Congratulations, Lara! It is fitting that you should have a fanboy
> such as gavino.
That's one more than you have.
Your postings, your Ruby-code and your opinions are totally
uninteresting, dull and insignificant; a mere annoyance.
You have _nothing_.
--
Lars Rune Nøstdal || AJAX/Comet GUI type stuff for Common Lisp
http://nostdal.org/ || http://groups.google.com/group/symbolicweb