English version
Accueil     À propos     Téléchargement     Ressources     Contactez-nous    

Ce site est rarement mis à jour. Pour les informations les plus récentes, rendez-vous sur le nouveau site OCaml à l'adresse ocaml.org.

Browse thread
[ Home ] [ Index: by date | by threads ]
[ Search: ]

[ Message by date: previous | next ] [ Message in thread: previous | next ] [ Thread: previous | next ]
Date: 2007-08-31 (22:16)
From: Richard Jones <rich@a...>
Subject: virt-top

Virt-top is a practical measurement tool for monitoring the
performance of virtual machines.  Unix users can think 'top' for Xen,
QEMU, etc.

At some point I want to write a paper about my experiences with
writing this in OCaml, but in brief:


* I've had several comments about how easy the binary is to install --
it doesn't require lots of runtime infrastructure like an equivalent
Python program might.  It's just a native binary which is dynamically
linked to libvirt.so & ncurses.so.  People who aren't paying close
attention probably think I've written a C program.

* The program is incredibly short considering the amount of stuff
which it does -- just 1000 LoC.

* The program takes a tiny fraction of processor / memory than people
are used to.  And it doesn't leak memory.  (Believe me, these are
baffling new "features" compared to similar Python programs).

* Not having to worry about the usual C hassles (memory management,
strings, errors etc.) makes programming much simpler.  I know because
today I mistakenly started another project using C, and realised what
a grind it was -- a quick rewrite of that one in OCaml was necessary.

And now for the CONS:

* I have to provide binaries for most people, because compiling the
program is practically impossible for most users.

* Porting it to new platforms is hard, because of lack of
comprehensive OCaml support on things like FreeBSD / Solaris/ Gentoo /...
[I'm working on rectifying this and the above, but could do with help
from any FreeBSD and _especially_ Solaris users out there].

* Lack of library support is so acute that I've had to structure the
program in several "layers" so that, for example, only extlib &
ocaml-curses is absolutely required (it'd be hard to imagine a
top-like program which didn't use ncurses), but users can still build
from source with other OCaml libraries missing, although they will get
reduced functionality.
[See: http://hg.et.redhat.com/virt/applications/virt-top--devel?f=81bcc3563dc7;file=virt-top/README]

* Most people look at the code in confusion.  I still haven't had a
single patch from anyone, and that is disappointing (and more
importantly, against the kind of open involvement which Red Hat

* Tools: Common text editors don't know what to do with the .ml files.
Even emacs doesn't have tuareg-mode by default.  This extends
throughout the whole toolchain.  For example source control tools
can't annotate the code
(http://hg.et.redhat.com/virt/applications/virt-top--devel?f=2f31261c9437;file=virt-top/virt_top.ml - q.v.),
and automake doesn't understand what to do with .ml files.

So do we want to make functional programming available to the masses?
It's all the "stupid" stuff like the above which needs the effort to


Richard Jones
Red Hat