HiBuffer: Buffer Analysis of 10-Million-Scale Spatial Data in Real Time
Round 1
Reviewer 1 Report
The paper demonstrate an end-to-end web based system for buffer analysis/visualization using parallel computing. The paper is clearly written and has provided extensive experiment results.
Pros:
end-to-end fully parallelized system for buffer analysis
tiled-pyramid for visualization
Cons:
in the experiment section, the setup is a single node machine which cannot demonstrate the uses of MPI for scalability. To my understanding, MPI is used as a task tool and OpenMP is used as a thread pool, which might not fit for the primary use case of MPI and OpenMP. It makes more sense if the setup is a cluster rather than a single node.
the datasets in experiments seems not very big and the their R-tree indexes are small. what if the dataset is very big and R-tree index is also not small. How would the R-tree index hub be like?
it seems that the pixel based approach is very similar to rendering in computer vision. what is the difference comparing against CV methods? how does this system compare with GPU based rendering?
Author Response
Dear reviewer,
Thanks for your valuable comments and suggestions, which are very helpful in improving the quality of our work. All the comments have been addressed as follows. Each comment has been quoted in italics while the revisions are marked in red in the manuscript.
Author Response File: 
Author Response.docx
Reviewer 2 Report
This is a great paper! The authors presented a very interesting real-time buffer tool (HiBuffer) which is able to generate buffers for large-scale spatial data sets very quickly. In general, the paper is well written and easy to follow. The authors have validated the performance of the proposed tool with several real-world data sets and provided in-depth sensitivity analyses of the significantly-improved performance of HiBuffer (compared to other commonly-used GIS software tools). I only have a few minor comments. The paper can be accepted for publication after these comments are properly addressed.
1. In the Spatial-Index-Based Buffer Generation method, it is clear that the target point to be detected (i.e., P in Figure 4) are not within the buffering zones of any points/lines located outside of the outer box; meanwhile, the target point is obviously within the buffering zones of any points/lines which are located inside of the inner box. However, it is not clear how do you handle any points within the area the between the inner and the outer boxes? In fact, detecting these points is more challenging than the points beyond the outer box or within the inner box. The author should also take this into account in Algorithm 1.
2. The authors introduced the commonly-used tile-pyramid approach to help generate real-time buffering zones for large-scale data sets. However, the tile-pyramid approach is essentially based on some tiles which are not generated beforehand and thus are not truly real-time (This is especially true for raster data sets). In your method, do you create all the buffer tiles beforehand? Or do you create each tile in a true real-time manner? In other words, if I want to explore the buffer zones for the same region with the same zooming level for many times, does your tool create the buffer every time? Or it creates all buffer tiles for this region beforehand and then loads the corresponding tile from the storage pool every time? Please clarify with more details.
3. The authors have tested the tool with a SMP server. How about its performance in a personal computer?
4. Table 2: It seems you only consider line and point features for testing the buffer zones generated from the proposed tool. How about polygon features? How does your tool deal with polygon features?
5. For Section 5 – Online Demo of HiBuffer, it is suggested to include a screenshot of the online demo in the paper.
6. Page 13, lines 325-326: change “for it takes” to “because it takes”; change “for this setting has” to “as this setting has”.
Author Response
Dear reviewer,
Thanks for your valuable comments and suggestions, which are helpful in improving the quality of our work. And we are very happy for your approval on our work. All the comments have been addressed as follows. Each comment has been quoted in italics while the revisions are marked in red in the manuscript.
Author Response File: Author Response.docx
