Optimizing allocation of on-demand resources using performance

1. A system comprising:

an access layer having a plurality of server collections, each including a plurality of servers;

an aggregation layer having a plurality of first switching devices each coupled to a portion of the plurality of server collections to enable communications with ones of the plurality of server collections coupled to the corresponding first switching device;

a core layer including a plurality of second switching devices, each coupled to a portion of the plurality of first switching devices, to enable communications with others of the plurality of server collections not connected to the corresponding first switching device;

a plurality of performance zones, each performance zone formed of and corresponding to a logical collection of the portion of the plurality of server collections coupled to the same first switching device, wherein communication within a performance zone is optimized as compared to communications between performance zones;

a single logical network for a first customer, the single logical network formed of at least some of the plurality of performance zones;

a logical aggregation layer to logically couple a first performance zone and a second performance zone, wherein the first performance zone comprises a full mesh network including a first plurality of servers each logically coupled to each of the others of the first plurality of servers via a virtual switch;

a scheduler to dynamically allocate an on-demand resource requested by the first customer on one of the first plurality of servers, wherein the first customer is associated with a performance zone identifier for the first performance zone, wherein the on-demand resource is allocated with a quality of service guarantee when the on-demand resource is allocated within the first performance zone and without a quality of service guarantee when the on-demand resource is not allocated within the first performance zone; and

a logical switch having a table for the first customer including a plurality of entries, each of which associates an on-demand resource of the first customer to a performance zone identifier corresponding to the given performance zone, wherein the logical switch enables the first customer to view a plurality of on-demand resources of the first customer located in different performance zones as the single logical network to abstract the performance zones away from the first customer.

2. The system of claim 1 , wherein the scheduler is to weight a request by the first customer for allocation within the first performance zone with a first value and to weight a second allocation criteria with a second value.

3. The system of claim 2 , wherein the scheduler is to schedule the on-demand resource to the first performance zone based on the weighted allocation request and the weighted second allocation criteria.

4. The system of claim 1 , wherein the on-demand resource comprises a database resource within the first performance zone or a storage resource coupled to the first performance zone via the aggregation layer and the core layer.

5. The system of claim 1 , wherein each performance zone is a constrained logical portion of the single logical network.

6. The system of claim 1 , further comprising a logic to trigger an event to create a network reservation via at least one tunnel when the first customer provisions a second resource in the second performance zone, wherein the first customer previously provisioned a first resource in the first performance zone.

7. The system of claim 6 , wherein the network reservation is of a predetermined bandwidth.

8. The system of claim 1 , wherein when the on-demand resource is not allocated within the first performance zone, a tunnel having a first amount of bandwidth is dedicated between the on-demand resource and other on-demand resources within the second performance zone.