09/07/2004 6:47 PM
post4463
|
Comments for the OGSA V1
First of all, congratulations to all of you! I've been a witness (spectator) of how hard and long have you worked on
this. Secondly, my comments have been triggered by reading something that was not clear for me or because I felt that
something was missed on the descriptions or that some concepts were not enough clear for a specification document.
Following are my comments:
1) Page 8, on top: 2.1 Dynamic and Heterogeneous...: Heterogeneous systems should establish and publish or make known
their requirements on how to allow access to their resources.
2) page 9, on top, 2.3 Optimization: Q: What is Grid workload? From a grid viewpoint the "workload" can be in the
traffic but the real workload and SL objectives should depend on how the workload is distributed and how each node
schedules and manages its own workload (as it is more or less said in 2.5 Job Execution).
3) page 10, 2.6 Data Services, Data access: The heterogeneous system should provide the necessary interfaces to allow
data access to a request coming from a different system.
4) page 12, 2.10, Availability (Don't know where to say it): OGSA should require to the service providers a log or, at
least, the status of the offered service.
5) page 12, 2.11 Ease of Use...: (Perhaps because my english):Q: Why, in this particular case, the complexity should be
masked?
6) Page 13, Figure 1, 2nd and 3rd layers: I don't see a strong "barrier" between these two layers, mainly in the case of
User Domain Applications and User Frameworks (3rd layer), and Security Framework and SLA Management (2nd layer)
where necessarily must occur a closed interrelationship at least to gain access to the grid.
7) Page 18, 3.4 Execution Management Services: Q: What is a non-legacy batch job?
8) Page 18, 3.4.1 Objectives: Too many unanswered questions. Specs should provide an answer (tacit or explicit) to
everyone.
9) Page 21, 3.4.5.1 Job: The three paragraphs are confusing and in some aspect contradictory with what is said in 1st
and following paragraphs of 3.4.5.2, and in 3.8.3.1 where a job seems to be what (normally) is understood as a job.
10) Page 24, 3.4.8.1 Case 1..., 2nd paragraph: Q: Does the JM know if where "he wants to run the service" has other
installed software that are incompatible with this new version?
11) Page 27, 3.5.2.1 Types of Data Resource: Q: Are the simple record-file (e.g. record length equal 1000 bytes)
included in Flat Files or in Streams?
12) Page 50 and following, 3.9.2-3.9.6: Be aware that has been introduced the terms "producer" ( also "or intermediary")
and "consumer" instead of the WS' roles of provider and requestor.
End of my comments. Again my congratulations and good luck!. Olegario Hernandez, e-mail address: oleghf@vtr.net
|
|
|