CS848 Paper Review Form - Fall 2006 Paper Title: Database Replication Policies for Dynamic Content Applications Author(s): Gokul Soundararajan, Cristiana Amza, Ashvin Goel 1) Is the paper technically correct? [X] Yes [ ] Mostly (minor flaws, but mostly solid) [ ] No 2) Originality [ ] Very good (very novel, trailblazing work) [X] Good [ ] Marginal (very incremental) [ ] Poor (little or nothing that is new) 3) Technical Depth [ ] Very good (comparable to best conference papers) [X] Good (comparable to typical conference papers) [ ] Marginal depth [ ] Little or no depth 4) Impact/Significance [ ] Very significant [X] Significant [ ] Marginal significance. [ ] Little or no significance. 5) Presentation [ ] Very well written [X] Generally well written [ ] Readable [ ] Needs considerable work [ ] Unacceptably bad 6) Overall Rating [ ] Strong accept (very high quality) [X] Accept (high quality - would argue for acceptance) [ ] Weak Accept (marginal, willing to accept but wouldn't argue for it) [ ] Weak Reject (marginal, probably reject) [ ] Reject (would argue for rejection) 7) Summary of the paper's main contribution and rationale for your recommendation. (1-2 paragraphs) This paper provides an alternative solution to the back end database tier of the multi-tiered web application: rather than using expensive centralized hardware, people can use database replication over a cluster of relative cheaper servers. The advantage of the solution is not only more economical, but also data replication can balance the workload of each server and provide good fault tolerance. In addition, the mechanism presented in this paper can also dynamically adjust the number of replicas for each application to maintain application-level performance. Furthermore, this approach requires no modification to web-server and the web-application, so it is easy to deploy. 8) List 1-3 strengths of the paper. (1-2 sentences each, identified as S1, S2, S3.) S1: the paper uses a mapping policy called partial overlap that provides fast data migration and minimizes interference across applications. S2: All the implementation of this approach is in resource manager and query scheduler, so there is no need to modify the existing web-server and web-application. 9) List 1-3 weaknesses of the paper (1-2 sentences each, identified as W1, W2, W3.) W1: When describing the replica allocation policy, it is not quite clear how to set up the values of HighSLAThreshold and LowSLAThreshold. W2: When the entire system is overloaded, the approach proposed in the paper will allocate same number of replicas for each web-application. It is good if customer can have an option to choose whether to allocate same number of replicas for every application, or give more important application more resource when system is experiencing overload. 10) Detailed comments for authors. Just a little concern: If the entire system is underload, based on this algorithm it wonĄŻt allocate more replicas for the application. Therefore there might be some servers in the cluster are unused. Is it a good idea to allocate the unused servers to some application so we can have better application-level performance?