[LinuxFailSafe] Contributing

[Lars Marowsky-Bree]

On 2002-10-10T12:17:22,
   Kashif Shaikh <kshaikh@consensys.com> said:

> Yeah, FailSafe tried doing everything...even tried to create something
> like a clustered filesystem(the cdb).

No, the CDB is actually a very cool piece of code, albeit implemented in
violation of the layering they intended to use orginally ;-)

But still, it is a distributed configuration "registry", which avoids one of
the most common errors - namely, configuring parts of the cluster differently
by cut&paste errors etc. This is actually one of the pieces I want to rescue
over to the new framework one day.

> But personally, I would like that Linux become distributed, so that
> low-level details like node memberships, synchronization primitives,
> ipc-communication(i.e. distributed shared memory) are actually part of
> the OS, and the end-user(programmer) will have a singular logical view
> of the OS. This won't only benefit HA clusters, but HPCs or
> load-balanced servers.

The so-called "Single System Image" clustering you are referring to here is
addressed by Compaq's SSI work on Linux, which is also Open Source. It
requires fairly drastic changes to the Linux kernel and it is rather difficult
to scale well (ie, distributing the load which was designed for UP/SMP style
systems). Some of the issues are discussed in "In Search of clusters", a real
classic.


Sincerely,
    Lars Marowsky-Brée <lmb@suse.de>

-- 
Principal Squirrel
Research and Development, SuSE Linux AG
 
``Immortality is an adequate definition of high availability for me.''
	--- Gregory F. Pfister