Talk:Multiple Domains
Revision as of 09:15, 20 October 2007 by Morgaine Dinova (talk | contribs) (→Adding an Organization Domain)
Adding an Organization Domain
- The addition of Organization Domains into this particular architectural view does not seem to be correct. Agent Domains provide CPU resources to power agents, and Region Domains provide CPU and storage resources to power regions. In other words, Domains are resource pools, and could quite correctly be termed "Resource Domains".
- In contrast, an "Organization Domain" is a type of "Authority Domain" and has nothing to do with object resourcing. If the architecture were multilayered then there would be room for an Organization Domain as a superdomain enclosing its own controlled Agent and Region domains, but in a single-tier architecture the Organization Domain shouldn't appear in this picture. It's orthogonal to Agent and Region domains. --Morgaine Dinova 04:44, 24 September 2007 (PDT)
- To put it another way, Organization is an attribute that applies to resource domains like Agent and Region. --Morgaine Dinova 05:20, 25 September 2007 (PDT)
How to link domains
- In trying to understand the basis for claimed scalability in this section, it struck me that it is hard to discern what is actually being depicted or explained in these diagrams. Obviously when you see two domains where there was previously only one then the implication is that the system can now handle more than one of them, but that is unrelated to scalability per se.
- Scalability is about avoidance of bottlenecks, not about how many things can be reached, and this was not addressed here I think. This is the primary task when designing for scalability, and not a side issue, so it's important that the items in the scalability viewpoint be highlighted and addressed. For example, diagrams for illustrating that viewpoint should show the datapaths and contention points and the resource provision. Legends proclaiming "It scales", "It scales" alone are not enough.
- I'm sure that this information is not detailed in the architecture yet simply because it is very early days, and it takes time to document, and the information was probably made plain to everyone verbally in the AWG meeting. This is fine, and patience is a virtue, but I do want to state that each of the dimensions that is being scaled has its own description needs, beyond mere ennumeration, and that is not coming across yet. On the plus side, I am very patient. :-)
- I think there needs to be a more explicit section related to scalability in the AWG wiki tree, seeing as scaling is such a fundamental part of this forward-looking project. Comments and suggestions please. --Morgaine Dinova 11:49, 26 September 2007 (PDT)