CS848 Paper Review Form - Fall 2006

Paper Title: Database replication policies for dynamic content applications
Author(s):Gokul Soundararajan, Cristiana Amza and 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
 [ ] Significant
 [x] 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)
 [ ] Accept (high quality - would argue for acceptance)
 [x] 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 presents a set of resource allocation poilicies to optimize 
performance of a shared cluster running multiple applications. The main idea 
is to allocate database servers from a pool to applications based on their 
workloads. The paper proposes two improvements in the basic idea. First, 
using delay-aware heuristics to prevent unwanted oscillation in the allocation.
Second, using a partial overlap mapping of replicas. In this mapping the read 
replica sets among applications are kept disjoint to avoid buffer cache 
conflicts. But keep additional write replica sets are maintaintained that may 
intersect among the applications in order to speed up the migration of the 
replica from one application to another when load varies.

The main use of such resource allocation scheme is data centres with hundreds
(or even more) of servers running numerous applications with different mix of
workloads. But this techinique is not going to scale well for number of replicas
significantly more than 8, as shown in figure 2 and 3 in the paper. The 
applicability of the scheme seems to be limited and insignificant. Hence the 
given rating.


8) List 1-3 strengths of the paper.  (1-2 sentences each,
identified as S1, S2, S3.)

S1: Unified approach to resource allocation and fault tolerance
S2: Simplicity

9) List 1-3 weaknesses of the paper (1-2 sentences each,
identified as W1, W2, W3.)

W1: This technique is not going to scale for number of replicas
    significantly more than 8, because of the write-all read-one
    replication scheme.
W2: Lots of graphs, but too little explanations.

10) Detailed comments for authors.

Requires more description and explanation for the scheme and and graphs.