CS848 Paper Review Form - Fall 2006

Paper Title: Replication and consistency: being lazy helps sometimes

Author(s): Yuri Breitbart and Henry F. Korth

1) Is the paper technically correct?
 [Y] Yes
 [ ] Mostly (minor flaws, but mostly solid)
 [ ] No

2) Originality
 [ ] Very good (very novel, trailblazing work)
 [Y] Good 
 [ ] Marginal (very incremental)
 [ ] Poor (little or nothing that is new)

3) Technical Depth
 [Y] Very good (comparable to best conference papers)
 [ ] Good (comparable to typical conference papers)
 [ ] Marginal depth
 [ ] Little or no depth

4) Impact/Significance
 [ ] Very significant
 [Y] Significant
 [ ] Marginal significance.
 [ ] Little or no significance.

5) Presentation
 [Y] Very well written
 [ ] Generally well written
 [ ] Readable
 [ ] Needs considerable work
 [ ] Unacceptably bad

6) Overall Rating
 [Y] Strong accept (very high quality)
 [ ] 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)
The paper proposes a new replica management protocol "Protocol GS" with
lower message overhead and asymptotically fewer deadlocks than 
previous results. The protocol works like a better "Lazy-master" and makes 
distinguish between local and global transactions. It proposes new concepts 
"virtual sites" and "replication graph" with several very handy theorems 
for global serialization and deadlock, simplifying the analysis of replica 
management. 

8) List 1-3 strengths of the paper.  (1-2 sentences each,
identified as S1, S2, S3.)
S1: Successfully reduce complex replica managment problems to graph notions,
    simplifying analysis. 
S2: Proper selection/construction of examples for different problem scenarios 
    and the conciseness and thoroughness of proofs in sound forms.
S3: Successfully developed a set of theorems for easy use instead of just 
    describing concepts, for example, theorem 1 can be easily applied in 
    analyzing global serialization.

9) List 1-3 weaknesses of the paper (1-2 sentences each,
identified as W1, W2, W3.)
W1: Lack of explicit elaboratioin of the rationale of locality rule, union 
    rule and split rule in virtual site management. In other words, for 
    example, the paper says nothing about why in locality rule, each local 
    transaction execute should be precicely ONE virtual sites, or why in 
    union rule, virtual sites at site Sj MUST be the same and MUST contain 
    all data at Sj accessed so far by Ti or by Tk. The paper just describes 
    them and says they are required but no explicit explanations about why. 
    This type of problems occurs in several other places in the paper which 
    concerns the integrity of argument in that if alternative factors are not
    excluded reasonably well, any possiblity of alternatives would compromise
    the foundation of optimality of what's proposed. 
W2: More performance evaluations, especially some prototype implementations with
    performance analysis demonstrating how this model can work/adapt in realworld 
    scenarios are expected. 
    
10) Detailed comments for authors.
The paper is a great job. It is the very kind of "best conference paper" with
not only scientific significance but also is concise and elegant for being able
to reach such depth and thoroughness in both deliverying and proving ideas 
systematically and mathmatically. A bit of improvement may come in adding 
performance evaluations and some prototype implemenations with realworld 
performance analysis demonstrating the feasibility, usability and compatibility
of the proposed model.