CS848 Paper Review Form - Fall 2006

Paper Title: Ganymed: Scalable replication for transactional web applications
Author(s): Christian Plattner and Gustavo Alonso


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

2) Originality
 [ ] Very good (very novel, trailblazing work)
 [ ] Good 
 [x] 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)

The paper provides a middleware-based solution for scalable and
consistent replication technique for database clusters to be used by
web applications. The replication technique proposed is an eager
primary copy approach with all the master objects residing in the same
replica. This is basically no different than passive replication. The
technique for selecting replica for read-only transactions is not
described in sufficient details. Replica selection in a passive
replication based on client requested response time and staleness
threshold is studied elaborately by Krishnamurthy et al. in the AQuA
project at University of Illinois during 2001-2002.

For a 2004 paper, there is no originality or novelty of the proposed
techniques except pointing out the fact that this approach actually
ensures a Snapshot Isolation consistency.


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

S1. A consistent replication technique without loosing scalability or
pre-determining the transactions patterns.

S2. Middleware based solution that hides the replication issues from
the clients and requires minimal modification in the database.

S3. Generally well written


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

W1. No significant technical contribution.

W2. No citation and comparison with the work by Krishnamurthy et al.,
which already achieved the same goal using a very similar approach.

W3. Clients with read-only transactions can either specify a staleness
threshold or the condition that client sees its own writes. "Read your
writes" is only one out of four different kinds of Session
Guarantees. More discussion was required in this issue. 


10) Detailed comments for authors.

C1. Krishnamurthy et al. of University of Illinois have published
several papers on passive replication in AQuA middleware during
2001-2002. They have also studied replica selection based on client
requested response time and staleness threshold elaborately. How do
the techniques used in Ganymed compares with those of AQuA?

C2. In couple of places in the paper it was mentioned that the
inconsistencies among the replicas at any time are hidden from the
client or read-only transactions will always see fresh data. This is
not entirely true since staleness in allowed in the secondary replicas
and clients can decide to access stale data.

C3. Why does a two-replica system perform poorly compared to both a
single and a three-replica system?