Hybrid reliability and cost-based method for resource allocations in software defined infrastructures

1. A computer-implemented method of allocating common resources for a user in a cloud computer network, comprising:

monitoring system reliability and resource charges in a monitoring component coupled to a client computer of the user and the resources;

first estimating a reliability based on the monitored system reliability and simulating resource provisioning choices to measure an impact of resource provisioning on the estimated reliability;

second estimating a cost based on the monitored resource charges;

allocating, in an allocation component managing the common resources, network resources based on the estimated cost and simulated resource provisioning choices;

maintaining a system reliability model to perform the first estimating and that is used to forecast when the system is at risk of instability or failure based on a change of one or more reliability measures, and wherein the reliability model utilizes a survival function based on workload stress introduced by each job of a plurality of jobs, wherein workload stress comprises a number of independently and identically distributed random variables that have either a discrete or continuous distribution; and

updating the system reliability model on a first periodic basis to adapt the reliability model to ongoing system changes.

2. The method of claim 1 wherein the common resources comprise at least one of: server resources, virtual machines, data containers, storage capacity, network interface resources.

3. The method of claim 1 further comprising:

maintaining a cost model to perform the second estimating; and

updating the cost model on a second periodic basis to adapt the cost model to the ongoing system changes.

4. The method of claim 3 wherein the user comprises a networked client sharing the common resources with a plurality of other clients through one or more processing queues.

5. The method of claim 4 wherein the system changes comprise at least one of: resource failures, changes in memory levels, and number of clients in processing queues.

6. The method of claim 3 wherein the costs comprise one of: a cost in revenue of the system failing to provide a service, and an estimated cost impact due to user dissatisfaction.

7. The method of claim 1 wherein the survival function allows for a nonconstant job arrival rate, any distribution of job service times, and any distribution of workload.

8. The method of claim 7 wherein the survival function comprises an alternative metric comprising one of: a hazard function, a failure probability, a cumulative hazard function, a mean residual life, a probability density function, a reliability function, and a mean time between failures measure.

9. The method of claim 1 wherein the survival function comprises a probability of a resource surviving past a given time, the method further comprising:

measuring a remedial cost of failure of the resource or a cost of correction to remedy non-survival of the resource; and

defining an acceptable risk of failure based on the probability and the remedial cost.

10. A system for allocating common resources for a user in a computer network, the common resources comprising at least one of: server resources, virtual machines, data containers, storage capacity, network interface resources, and the system comprising:

a non-transitory computer-readable medium containing computer-readable instructions, that when executed by a computer cause the computer to perform steps comprising:

monitoring system reliability and resource charges;

first estimating a reliability based on the monitored system reliability and simulating resource provisioning choices to measure an impact of resource provisioning on the estimated reliability, and second estimating a cost based on the monitored resource charges;

managing the common resources, and allocating the network resources based on the estimated cost and simulated resource provisioning choices;

maintaining a system reliability model to perform the first estimating, wherein the system reliability model is used to forecast when the system is at risk of instability or failure based on a change of one or more reliability measures, and utilizes a survival function based on workload stress introduced by each job of a plurality of jobs, wherein workload stress comprises a number of independently and identically distributed random variables that have either a discrete or continuous distribution;

updating the system reliability model on a first periodic basis to adapt the system reliability model to ongoing system changes;

maintaining a cost model to perform the second estimating; and

updating the cost model on a second periodic basis to adapt the cost model to the ongoing system changes.

11. The system of claim 10 wherein the user comprises a networked client sharing the common resources with a plurality of other clients through one or more processing queues, and wherein the ongoing system changes comprise at least one of: resource failures, changes in memory levels, and a number of clients in processing queues.

12. The system of claim 10 wherein the survival function comprises a probability of a resource surviving past a given time, and the steps further comprising measuring a remedial cost of failure of the resource or a cost of correction to remedy non-survival of the resource, and defining an acceptable risk of failure based on the probability and the remedial cost, wherein the costs comprise one of: a cost in revenue of the system failing to provide a service, and an estimated cost impact due to user dissatisfaction.