On Tue, 2005-06-28 at 15:39 +0100, Gordon, JC (John) wrote:
> Andrew, there is undoubtedly research happening but here we are trying
> to get a production infrastructure in place, not a research project.
These two classifications are not mutually exclusive. Certainly from
what I've seen first-hand, LCG is still undergoing the rapid change more
assosciated with a project still under development than a production
release receiving smaller service increments. It has yet to really
settle down into a stable distribution.
> The foundation is not dCache but SRM so we are not locking ourselves into
> dCache as there will be alternatives (we know they are under development).
Okay.
> For the moment however, dCache appears to be the only
> implementation anywhere near production quality. DPM is designed for
> small sites with a few disk servers; it doesn't claim to be designed for
> MSS systems so dCache is the only player in the short term.
For sites with large amounts of storage, perhaps -- but some of the
sites involved _are_ small sites with only a few disk servers; I would
suggest that DPM may be a better solution to deploy in some of these
cases.
> If adopting dCache for now means sites lose the ability to bend the code
> to their will then that is a price I am willing to pay for a production
> service. It is not as if dCache comes from some anonymous corporation,
> we know the developers and talk to them.
And they've said "we're not going to publish the source." And that's
fine; they're free to do that. However, given that we'd like to have
fully open source solutions, the use of dCache should be marked as a
deprecated option.
This is equivalent to saying "We're going to be using this tool for the
time being, but we're planning to change to something else at some point
in the future as new options become available".
Cheers,
David
--
David McBride <[log in to unmask]>
Department of Computing, Imperial College, London
|