Nxt-Freedom: Considering VDC-Based Fairness in Enforcing Bandwidth Guarantees in Cloud Datacenterâ€
Round 1
Reviewer 1 Report
The paper investigates bandwidth allocation in a multi-tenant data center. The paper first propose a model for the data center tenants to specify their network requirements called Switch Centered Graphs, SCG. Then the paper proposes a bandwidth allocation with the following characteristics:
- Fair allocation, where redundant bandwidth is evenly distributed among different tenants
- Work conserving (no link bandwidth resources are wasted)
- Flow based bandwidth enforcement, where a congestion controller decides the proper rate of every flow.
- Flows bandwidth are computed solving the maximum flow problem using the well known Edmonds-Karp algorithm
Paper strengths:
- The research topic is interesting and timely.
- The research topic is well motivated in the abstract and introduction.
- The paper is well structured and readable.
- The proposed models are sufficiently explained and described.
Paper weaknesses
- The research topic is not new: other network abstractions to describe tenants network requirements, and bandwidth allocation algorithms have been proposed in the literature.
- Although the authors point out some advantages of their proposal over others found in the literature, they do not mention features where those proposals might be better. For instance, in reference [25] of the paper, the bandwidth reservation scheme is targeted to time-varying network reservations.
- The authors claim as the main contributions of their work that their bandwidth reservation scheme performs fair allocation and is work conserving. However, there are some recent works not mentioned in the paper that address these issues too, as the references below:
- B. S. Ali, K. Chen and I. Khan, "Towards Efficient, Work-Conserving, and Fair Bandwidth Guarantee in Cloud Datacenters," in IEEE Access, vol. 7, pp. 109134-109150, 2019, doi: 10.1109/ACCESS.2019.2930888.
- Chen, Kang, and Ning Yang. "BwShare: Efficient bandwidth guarantee in cloud with transparent share adaptation." Computer Networks 170 (2020): 107095. https://doi.org/10.1016/j.comnet.2019.107095
- The evaluation scenario is very limited. The authors use a virtualization environment with only 2 tenants having respectively 4 and 2 VMs.
- Efficiently of the proposed bandwidth reservation algorithm under time-varying network reservations is not analyzed.
Other comments:
- The bandwidth allocation is done on per flow basis. It would help the reader if the paper define "flow". I guess it is the 4-tuple formed by the IP address and ports.
- It is not clear to me why bandwidth allocation is done on per flow basis. Why not enforcing bandwidth allocation on VM-source to VM-destination basis? There can be hundreds of TCP connections (flows) between VMs, thus, impacting scalability.
Minor comments:
- Algorithm 1, page 12 is not referenced in the paper
- reference 12: Technical report, 2000. -> RFC2992?
- reference 31: the link is not reachable
Typos:
line 40: way. in -> way. In
line 42: ...hypervisor are employed, which. -> broken line
line 118: ofcongestion -> of congestion
line 218: The endpoint link non-blocking For arbitrary -> The endpoint link non-blocking: For arbitrary
line 319: indicat -> indicate
line 374: is introduce -> is introduced
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper presents an interesting study on fair resource allocation in virtual datacenters. The authors present algorithms for fair allocation of resources. The paper is generally well written except for few minor typos and grammar mistakes here and there.
The algorithms presented rely on a number of assumptions. It would be good that the authors provide some comments on the practicality of those assumptions.
There are also a number of undefined acronyms in the manuscript including the abstract. Please ensure all acronyms are defined at first use for clarity.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The topic of the paper looks interesting. Congratulation on your efforts. The experimental results are impressive.
I have a few suggestion to improve the quality of the writing.
- Rephrasing the definition of fair redundant bandwidth and Work-conserving would help the readers to understand them better.
- The abbreviation SCG is used on line no 138 but its full form is given at 163. It would be better to use the full form at the first occurrence.
- At the line number 195, is it a virtual datacenter or virtually?
- At line number 199, I am confused on whether it is number of ports or port number.
- In section 3.3 at page 6, the terms blocking and non blocking are used first time. It would be helpful for the reader if the terms are explained there instead of explaining them later in page 12 .
- Also the statement before the equation1 between the lines 218 and 219 needs grammatical check.
- it would be better to refer the equations in the text to provide a connection.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
