The use of a monolithic architecture become more and more difficult when the number of users and the number of resources grow simultaneously. When a user tries to resolve a problem, without the multi-MA extension, DIET looks for the better SeD that can solve it. This search involves the fact that each SeD has to be queried to run a performance prediction as described in Section 1.4.
The need to query every SeD that can resolve a problem is a serious scalability issue. To avoid it, the multi-MA extension proposes to interconnect several MA together. So, instead of having the whole set of SeD available under a hierarchy of a unique MA, there are several MA and each MA manages a subset of SeD s. Those MA are interconnected in a way that they can share the access to their SeD s.
Each MA works like the usual: when they received a query from a user, they looks for the best SeD which can resolve their problem inside their hierarchy. If there is no SeD available in its hierarchy, the queried MA forwards the query to another MA to find a SeD that can be used by its client. This way, DIET is able to support more clients and more servers because each client request is forwarded to a number of SeD s that is independent of the total number of available SeD s.