Re: Why i??? NEED YOUR INPUT

Aaron,

It's encouraging to see you thinking about cloud computing under IBM i. It sounds like you have a pretty good handle on the types of things needed to set up separate environments for separate organizations, without partitioning. One idea would be to allow users to set up any number of libraries themselves, but have them use a library name prefix unique to each - so that all their library names begin with their prefix.

I might look into the possibility of different organizations running applications under separate subsystems, including their HTTP server jobs - don't know if that's possible. Maybe move system commands out of QSYS, and allowing users to override them with modified default values, using commands in their libraries. Naming convention for user profiles? Separate output queues. There's probably more.

-Nathan.





----- Original Message ----
From: Aaron Bartell <aaronbartell@xxxxxxxxx>
To: Midrange Systems Technical Discussion <midrange-l@xxxxxxxxxxxx>
Sent: Thu, February 25, 2010 3:58:40 PM
Subject: Re: Why i??? NEED YOUR INPUT

You can install your applications in one library, while setting up other

libraries for data - separated by organization. It's not like you have to
deploy an application runtime environment under separate partitions.

Very true. Just thinking off the top of my head (vs. the bottom :-), here
are the things a development environment within an LPAR would need to
facilitate for an RPGCGI programmer:

- Access to two libraries - one for the CGI programs to run in and be
exposed to the outside world, and one for other programs that you don't want
invoked as CGI and DB2 tables.

- Access to an IFS folder where they could add new files (.html, .jpg, .js,
.css, etc)

- Ability to start/stop their Apache instance

- Ability to view the logs of their Apache instance

- Ability to view the spool files generated by their Apache instance

A program could be written to quickly setup all the above for a new user
with a few key strokes (I have most of that code already written). The
access to active jobs and spool files could use adoptive authority as long
as the profile accessing the job matched the server instance name (i.e. we
would name profiles USR0001 with a lib of USR0001 and an Apache instance of
USR0001). Am I missing anything?

I would really like to get this infrastructure off the ground so people can
start trying out RPGCGI programming on a system that is all setup for them.

Aaron Bartell
http://mowyourlawn.com
http://mowyourlawn.com/blog/