A Fast Cold-Start Solution: Container Space Reuse Based on Resource Isolation

Round 1

Reviewer 1 Report

The paper submitted for review, titled A Fast Cold-start Solution：Container Space Reuse Based on Resource Isolation, addresses a very important and timely topic. In recent years, in tandem with the development of measurement systems and the use of the global network for collaboration, more and more information is stored in digital form. Unfortunately, the dynamic increase in the amount of stored information makes it difficult to use it in practice. An additional very important problem is the escalation of privileges or disclosure of information. Thus, it is expedient to take measures to improve and secure the processes related to their practical use.
The work is prepared very carefully. It contains sufficient introductory information on the subject. Most of the publications are from recent years. I suggest a more precise formulation of the purpose of the work. The way the work is carried out and the presentation of the results is very good. The only criticism is that the conclusions are very short. I believe that the research done allows for more elaborate conclusions.
Specific comments on the work:
Line 134 - I propose to characterize all the chapters of the work.
Line 178 and 219 - Why the larger font?
Figure 8 - What is the reason for the sudden increase in cold start delay time for selected functions.
Have you analyzed the energy inputs for the operation of individual systems? If so, what are the conclusions.
Very poor reference of the content of the paper to the conclusion: If the Alphidae alleviates the cold-start of functions, the time will be reduced from hundreds of milli-420 seconds to 18 milliseconds for general function calls.

Author Response

Please see the attachment

Author Response File: Author Response.pdf

Reviewer 2 Report

The authors of the text deal with the area of computational methods and tools in the field of distributed computing and simulation. With the advent of cloud-native computing, serverless computing has become interesting due to its scalability and flexibility and is being deployed for computationally intensive applications especially in the area of big data processing on service offering platforms. Serverless computing has an attractive application trend for a growing number of Internet services. The question is how to more efficiently address container resource utilization and service startup times for serverless.
The authors have focused their research from the assumptions of the complete life cycle of serverless functions, and they are concerned with improving the performance of serverless functions by changing the original methods of exchanging space for time or time for space. They focus on how to reduce the cold start time of a serverless system while maximizing the utilization of container resources. As a novelty, the design of dynamic control functions and container pools mainly involves three aspects. First, we create a container pool with classification identification based on the function utilization rate. Then, we use namespace technology to achieve container resource reuse in a security-isolated state. Next, we adaptively adjust the correspondence between functions and reusable container resources by monitoring system resources.
Finally, the test results show that by transferring the remaining container space resources to the container in the preheated state for a new function, the resource wastage caused by the function change can be effectively reduced. Idle function containers and container resource utilization can further reduce the cold start time while ensuring the safety and isolation of the function.
The paper is systematically and logisdically constructed, dealing with the current topic. I would recommend highlighting and delineating elements of novelty relative to previously known or published solutions. Formally, I have no major comments.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

In the abstract, I found the following issues/comments as

·       The problem needs to be clearly stated in the abstract.

·       The achieved experimental results are not appropriately described (as should be in %).

In introduction

·       The problem statement and motivation are not clearly stated.

·       The contribution in the paper needs to be revised, as these are not contributions.

·       The paper organisation paragraph needs to be included in the introduction section, not related work.

In related work, a detailed comparison table must be provided. 

·       I suggest the authors add recent publications related to a paper domain to strengthen your problem and arguments. For example, below are the few recent papers

 

o   Performance and isolation analysis of RunC, gVisor and Kata Containers runtimes. 

o   Opportunities for Optimizing the Container Runtime.

o   HRFP: Highly Relevant Frequent Patterns-Based Prefetching and Caching Algorithms for Distributed File Systems

o   Multiple linear regression-based energy-aware resource allocation in the Fog computing environment

o   Multi-criteria--based Dynamic User Behaviour--aware Resource Allocation in Fog Computing

o   Mobility-Based Resource Allocation and Provisioning in Fog and Edge Computing Paradigms: Review, Challenges, and Future Directions

o   Containerization technologies: Taxonomies, applications and challenges.

o   A micro-level compensation-based cost model for resource allocation in a fog environment

o   An efficient resource monitoring service for fog computing environments 

·       A detailed discussion section is needed before concluding the paper.

·       The conclusion section needs to be revised/modified with the achieved results.

 

A proof-reading of draft is required before submission for final draft. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

The authors have revised the paper according to all suggested feedback given in first review.