On 10/24/2012 10:12 PM, Matt Wagner wrote:
It feels a bit like to me like we're reimplementing half of
Conductor.
We're implementing the "good parts" and leaving the confusing admin
scraps behind. I understand the appeal of starting fresh, but I'm not
sure it's really the way to go. Why can't we just make Conductor more
useful for these "simpler" use cases? We used to have a self-service
portal but it kind of withered away. Why not just re-implement it,
following these ideas, and follow the good practices you propose to make
all of Conductor better and easier-to-use?
I'm not sure this is an either/or
choice. We can, and should continue to
make Conductor simpler to use, in all cases. Not least because Winged
Monkey is going to take time to implement, and longer to use a fully
featured Conductor API to expose Conductor's capabilities through a
simplified, end-user UI.
Until that day, Conductor has to continue to solve these problems
itself, so I think we should borrow good ideas from Jeremy's wireframes
etc., and use them to improve Conductor, wherever we can.
Worst case is that, in twelve months time, we'll have two applications
which both give a simple, elegant user experience. That would be a nice
problem to have.
At that point, it might make sense to de-scope some stuff from
Conductor, which, as mentioned already on this thread, attempts to do a
great many things, and could benefit from a simpler purpose, and to rely
on Winged Monkey instead. That's not a decision we have to face now though.
I also feel a bit like we're trying to make the app easy-to-use
by
leaving the hard parts out. Our configuration is indeed very complicated
and there's room for improvement, but I'm not sure that the fix for this
is to make a new app that ignores it entirely -- someone's still going
to have to solve the configuration part.
Solving the configuration part, doing all the hard parts, is the core of
what Conductor does well in its presentation of a managed, synthetic
cloud, as Hugh referred to it on this thread. Winged Monkey does indeed
have the luxury of ignoring some of those hard problems, but only
because Conductor has them covered.
One of the things that I think is attractive about Winged Monkey is
that, if it all comes together, it'll be possible to fire it up, and
give users a slick, intuitive interface, for a limited number of tasks,
without needing Conductor or Heat or I'd hope, even needing Deltacloud.
Winged Monkey could be pointed directly at ovirt/vsphere/openstack and
present a user portal.
Of course, some of the fancier features mocked up in Jeremy's wireframes
wouldn't appear in the Winged Monkey UI, but the core goal of allowing
users to manage their instances would work.
That's an attractive, versatile goal for Winged Monkey which has the
potential to make it interesting to several other projects.
The fact that Conductor has the account multiplexing, the multi-instance
deployables, the quotas etc., and we're working on features like cost
management means that Conductor is likely to enable more features in the
Winged Monkey UI than any other potential provider would but, going back
to the original point, that is because Conductor is doing a good job of
solving some hard problems.
But those questions/thoughts aside, I really like the Winged Monkey
idea. I think we've neglected the entire market of people that want to
manage a few instances, versus enterprise customers. As I said in
another thread today, I have a single persistent VM instance I use for
hosting my websites, and then I periodically spin up cloud instances to
test or tinker with things. Conductor is really not usable here -- I
don't want to set up all of the components, and I don't want to build my
own instances. I just want to manage a few instances, and launch
pre-built AMIs. (Not necessarily mine, though -- I just want to launch
the "Amazon Linux" AMI or whatnot.)
I'm also interested in some of the new features Winged Monkey seems to
add. It shows resource utilization and offers clone/snapshot options for
instances -- two things I was just wishing for in the earlier thread.
Looking at the wireframes, I also really like the way we express the
application lifecycle. The sparkcharts and other little charts are also
really slick and seem like they would work out really well.
I find it interesting that today, Mo sent out an email brainstorming a
neat idea for a cloud desktop client, I replied with a fairly different
idea that I thought would be more suited to a web app, and then you sent
out the information on what you folks have been working on for Winged
Monkey. Even though the ideas are a bit different, it seems like there's
a lot we have in common, and that it's something we've really got to do!
-- Matt