Message
The action items of the meeting:
(1) Hong will continue his focus on the implementation
of the proxy module.
(2) Aziz will continue his focus on the implementation
of the server module.
Estimated completion time: 2 weeks, after that, there
will be some integration work.
(3) Aziz will extend the project working plan draft
when more details become clear.
(4) Aziz will establish a Web page dedicated for
Sorrento. Hong will also contribute on that.
(5) Tao will send out a set of benchmarks that may be
used for Sorrento evaluation. Sezgin may contribute a distributed hash table
application, Lingkun may contribute a parallel protein sequence matching
program. However, bear in mind that the design of benchmark should take
advantage of the virtualization feature of Sorrento and seek to minimize
management effort while the set of cluster nodes change. (Questions you may want
to ask yourself is how do the data get initially populated to the system. What
happens when you want to expand the capability of the system. How do you expect
the storage subsystem to place data so that the performance will be the
best.)
(6) Jingyu will study PVFS, and run NPB (NAS Parallel
Benchmark) suite upon PVFS. (MPICH2 has support for PVFS). PVFS can be found at
http://www.parl.clemson.edu/pvfs/ I
had initially asked Bill to take a look at that problem. Maybe you both can work
together (Bill, can you forward the paper I sent you last time to Jingyu? That
paper contains a small section mentioning that they ran an NPB benchmark upon
PVFS).
(7) Jingyu, can you create a "sorrento" group on
Zaphod? Also please ask CS people to create a sorrento@cs.ucsb.edu email
account.
(8) There are not many independent pieces of components
remaining. However, everybody is welcome to checkout the code base through CVS
(CVSROOT is zaphod:/home/research_home/htang/cvsroot, project name cbs).
(9) One important component of the system which is open
for pick-up is the migration engine. The migration engine monitors the I/O
traffic on local data objects and determine whether a certain object may need
migration, and if so, to which other node.
(10) Nick has initially worked a little bit on the
kernel file system switch, tentatively Aziz is assigned the task of taking over
or collaborating with Nick on it.
Let me know if I missed something.
-Hong
Last modified: Fri Jun 6 01:13:25 PDT 2003