Review 1
Relevance and timeliness 	Technical content and scientific rigour 	Novelty and originality 	Quality of presentation
Good (4) 	Solid work of notable importance. (4) 	Some interesting ideas and results on a subject well investigated. (3) 	Well written. (4)

Strong aspects (Comments to the author: what are the strong aspects of the paper)
    This paper considers the problem of cloud control and the resulting framework unifies two other paradigms, load balancing and distributed rate limiting. The paper formulates the problem, proposes an algorithm to solve the problem and makes simulations to evaluate the behavior of the algorithm.
Weak aspects (Comments to the author: what are the weak aspects of the paper?)
    The algorithm presented in this paper has some restrictions and simplifications. The paper does not explain how the DNS mechanism interacts with the subgradient algorithm. Why is the communication between servers performed via UDP packets? TCP flows are more reliable and this communication is essential to the functioning of the proposed algorithm.
Recommended changes (Recommended changes. Please indicate any changes that should be made to the paper if accepted.)
    No recommendation.

Review 2
Relevance and timeliness 	Technical content and scientific rigour 	Novelty and originality 	Quality of presentation
Acceptable (3) 	Solid work of notable importance. (4) 	Some interesting ideas and results on a subject well investigated. (3) 	Readable, but revision is needed in some parts. (3)

Strong aspects (Comments to the author: what are the strong aspects of the paper)
    This paper focuses on the framework for usage control in cloud-based services. The main contributions of this work are following: a mechanism for cloud control that combines load balancing and distributed rate limiting; an algorithm for solving the optimization problem of interest is proposed based on standard sub-gradient method; and an analysis of the dynamical system modeling the per-server performance evolution is performed for a particular choice of performance metric. The paper is well formulated.
Weak aspects (Comments to the author: what are the weak aspects of the paper?)
    In section III, the evaluation is quite weak. There is no comparative analysis of the proposed mechanism with the existing mechanisms.
Recommended changes (Recommended changes. Please indicate any changes that should be made to the paper if accepted.)
    The reviewer suggests the authors to elaborate the performance evaluation of the proposed mechanism.

Review 3
Relevance and timeliness 	Technical content and scientific rigour 	Novelty and originality 	Quality of presentation
Good (4) 	Valid work but limited contribution. (3) 	Significant original work and novel results. (4) 	Well written. (4)

Strong aspects (Comments to the author: what are the strong aspects of the paper)
    The paper provides a novel idea for combining two problems facing management of cloud based services: load balancing and usage rate limiting while optimizing for low cost. Good analysis/solution of the optimization problem and convergence conditions.
Weak aspects (Comments to the author: what are the weak aspects of the paper?)
    Explanation about choosing an appropriate neighbor while updating the server capacities is not provided. Latency in communication between the peer servers is not considered. Results do not include responsiveness of algorithm. Simulation results could include more variations of dynamic workloads.
Recommended changes (Recommended changes. Please indicate any changes that should be made to the paper if accepted.)
    One/two typos found. Run spell check.

1 Meta review
Review 1
Relevance and timeliness 	Technical content and scientific rigour 	Novelty and originality 	Quality of presentation
Good (4) 	Solid work of notable importance. (4) 	Some interesting ideas and results on a subject well investigated. (3) 	Well written. (4)

Strong aspects (Comments to the author: what are the strong aspects of the paper)
    Interesting & relevant topic, good formulation & analysis
Weak aspects (Comments to the author: what are the weak aspects of the paper?)
    Proof of Thm. 1 not needed - more simulations with different D(y) functions would make the paper stronger.
Recommended changes (Recommended changes. Please indicate any changes that should be made to the paper if accepted.)
    What about "job affinity" (if jobs are classified into groups, include group type in decision to leverage caching, other locality-driven optimizations.