[coyotos-dev] Explicit Persistence Considered Harmful
Christopher Nelson
nadiasvertex at gmail.com
Mon Aug 18 17:17:53 CDT 2008
> 3. There are recoverability issues associated with implicit persistence.
> It is actually important that buggy program instances die and reload
> their state. In consequence, the argument that persistence eliminates
> the need for explicit storage seems dubious.
My thought on this is that there needs to be an ability to maintain
some sort of state that is not intimately bound to current runtime
state. Whether that happens to be an external filesystem, or
something else is another question entirely. Consider the example of
a webserver.
There is a certain amount of configuration data that needs to be
stored to initialize the server from a cold start. I can think of a
few ways of representing, for example, the httpd.conf sort of file.
(1) an external filesystem.
(2) a database server running either as cluster service (serving
several machines) or as a local service (servicing only the local
machine.)
(3) the ability to allocate persistent, non-volatile, named storage
that is rebound at runtime.
By (3) I mean that you allocate some memory space that is implicitly
persisted. However, if you have to cold start the process that
storage is not freed and reinitialized. Instead, it is rebound at
some predetermined place in the app's memory space, thus allowing the
app to reload its configuration.
One other thing to think about: providing persistent access to handles
to an "external" filesystem is probably inescapable, especially if you
intend to provide nfs or smb servers for coyotos storage.
-={C}=-
More information about the coyotos-dev
mailing list