A Virtual Network Resource Allocation Framework Based on SR-IOV
Round 1
Reviewer 1 Report
The authors propose a method to dynamically sense the network workload of a VM (using I/O para-virtualization, Single-Root-IO Virtualization and Virtual Functions) and make the resource scheduler to allocate resources accordingly. According to the authors, "the I/O characteristics of the VM network load cannot be obtained by the VMM, so that the network resources cannot be optimally scheduled". The authors propose to use the I/O para-virtualization for each VM, assign some specific Virtual Functions to network-intensive VMs. Their experiments show improvements in the network allocation, increasing the network throughput, reducing the delay of requests and speeding up the file download.
In chapter 2, the authors should explain better the para-virtualization and the Virtual Functions, since they are the core of the method proposed.
Chapter 3.1 lacks a diagram to help explain the basic idea and make the written text clear. Also, this text should be improved for clarity.
Figure 3 seems to lack some of the components referred in the text, making it difficult to follow the text.
In section 3.4.1, it should be discussed why it is difficult to obtain I/O load directly, instead of indirectly. In fact, the gathering of network I/O load is explained in section 3.4.2, so all this is seems to be poorly explained. In fact, by the end of chapter 3, it is explained how both i/O factors are used, but my question remains.
The sentence "We observed that I/O-intensive VMs have similar properties to I/O-intensive processes in Linux, that is, the CPU usage time of I/O-intensive VMs is always shorter, but it takes longer to wait for some I/O events to block" seems to be true, however it should be backed up by a reference or some data.
Formulas (1) and (2) should have the units used and all terms explained. For example, why is there a 10^9 in the formula?
In Chapter 4, table 2 should have depicted which VM is CPU-intensive (CPU-intensice?), network-intensive and mixed-load; and the %time in the columns.
Table 3 should have the results of the method applied. In Table 3, rows of VM1 and VM6 are the same. Is this a mistake?
Figure 6 is rather confusing. There is no reference for each of the four figures, and some legends seems to be not related with the chart.
General comments:
Overall I found the text confusing to read, namely the technical parts until chapter 3.3, making me go back and forth in the text to fully understand the meaning. This frustrates the reader.
The text in the figures is too small and many times there is a poor contrast with the background color.
The paper uses many acronyms that should be defined on first use.
The paper needs to be reviewed for english typos and clarity.
Author Response
The point-by-point response to the reviewer’s comments are included in the PDF file uploaded.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This paper introduces a method for sensing the workload measured in virtual instances with the aim to allocate resources on demand. Hence, enhancing the mechanisms to conduct the overall system performance are grounded on I/O para-virtualization and SR-IOV technology. The paper is easy to follow and readable but presents some areas of improvement that should be addressed before considering for publication. They are summarized as follows:
1. The contributions of this paper should be enumerated in the introductory section.
2. It would be appreciated if the authors provide more updated references in the state-of-the-art section when supporting the methods/algorithms used and assessed.
3. Section 3.1 (Basic Ideas) should clearly state the architectural principles, limitations, or relevant considerations towards the framework implementation.
4. Provide a more detailed representation of the System Framework. It is illustrated in Figure 3, but it is unclear to infer the information workflows. Enhance also the image resolution.
5. It is unclear the purpose of Figure 5. Only Fig. 5(a) has been referenced.
6. A more detailed description of the experimental testbed should be enhanced to ensure replicability
7. Section 5 (Related Work) is misplaced in the document. It seems to be allocated in the state-of-the-art section.
8. It is missing a discussion of the obtained results and its comparison with similar proposals.
Minor aspects:
- It would be appreciated to use more concise sentences in the text.
- Language spelling/grammar should be checked.
Author Response
The point-by-point response to the reviewer’s comments are included in the PDF file uploaded.
Author Response File: Author Response.pdf
Round 2
Reviewer 2 Report
In this reviewer's opinion, all the requested changes have been successfully implemented in the current version of this paper and it is ready for publication.
