CS848 Paper Review Form - Fall 2006
Paper Title: Facade: Virtual Storage devices with Performance Guarantees

Author(s): C. Lumb, A. Merchant, G. Alvarez

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
 [X] Very significant
 [ ] Significant
 [ ] Marginal significance.
 [ ] Little or no significance.

5) Presentation
 [X] Very well written
 [ ] 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)

The most important contribution of this paper is that it presents a 
technique through which performance guarantees can be provided on a 
storage system shared among various processes. It provides a framework
through which performance thresholds can not only be anticipated beforehand
but also specified, monitored and adhered to - even in extremely volatile
workloads.

My primary reason for accepting this paper is its practical signifiance.
The services of data centers and Storage Service Providers (SSP) are becoming 
more commoditized every day and SSPs need to multiplex their storage resources 
(which are usually large and indivisbile) for many small companies/applications. 
This paper addresses the problem of having some "fair" (where fairness is 
well defined in terms of an agreement) means of use in such  an environment 
and presents an elegant mechanism for providing fair service to users of a 
storage system - be they companies or individual users. 

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

S1 - Idea of managing queue depth and the feedback-control/scehduling algorithms,
in general were very simple. They were also and elegantly explained in the paper. 

S2 - The latency overhead induced by the virtual Facade Layer was negligible 
(appx 2%) making such a solution very feasible.

S3 - The test results were also well presented and covered a wide spectrum of
usage scneraios.

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

W1 - Although not directly a weakness there is no mention of how applications
will communicate the SLO with an IO request. The deadline, for example,
(part of the SLO) for each I/O request initiated by an application may be different.

W2 - SLO's are assumed to be static over time for a given stream and require 
administrator presence to be updated/inserted. This may not be a realistic 
assumption in many cases.

W3 - All physical disks are assumed to be similar in nature. 
Not a critical flaw but still worth noting that the assumption that small 
queue => better latency and large queue => better throghput may not
generalize well for Read only scenarios such as optical disks or linear 
magnetic tapes.

10) Detailed comments for authors.

- Generally well written paper.

- Layout of pictures and graphs needs some adjusting eg, reference to figures
several pages cauess flipping back and forth and is inconvinent.

- SLO's in the context of the paper was well defined but perhaps
some mention of how real user requirements in datacenters translate to these
lantecy SLOs would have been enlightening.