I've thought about it a bit, and strongly suggest first looking to see
if there is a reasonably good existing benchmark for JavaSpaces. If
there is, we should certainly consider it.
Otherwise, I suggest making the benchmark deal with two types of data,
"reference" and "transaction". Reference data is written rarely, remains
in the space for a long time, and is frequently read. Groups of
transaction entries are written by one transaction, and removed when it
completes. They have short lifetimes.
The benchmark should have parameters that let us shift proportions of
different entry behavior, numbers of threads, numbers of transactions etc.
I've written a relatively simple unit test for
com.sun.jini.outrigger.FastList that sometimes gets a null pointer
exception in it. FastList is the lowest level class involved in storing
outrigger's entries. As far as I can tell, my use of FastList conforms
to its interface, and the problem is internal to it.
Patricia
On 12/4/2010 5:50 PM, MICHAEL MCGRADY wrote:
I think you are more objective than that and would be interested in your
suggestion(s).
MG
On Dec 4, 2010, at 7:02 AM, Patricia Shanahan wrote:
Also, it may be worth searching for JavaSpace benchmarks. I don't want to write
or pick the benchmark myself, because it should reflect a users-eye-view of
JavaSpaces. I might accidentally bias the benchmark to reflect my current
thinking on performance improvement.
Patricia
On 12/4/2010 6:24 AM, Mike McGrady wrote:
Yes - there is. Let me noodle this. i will have to make adjustments, factor
in time, etc.
Sent from my iPhone
Michael McGrady
Principal investigator AF081_028 SBIR
Chief Architect
Topia Technology, Inc
Work 1.253.572.9712
Cel 1.253.720.3365
On Dec 3, 2010, at 9:34 PM, Patricia Shanahan<p...@acm.org> wrote:
MICHAEL MCGRADY wrote:
I might jump in here too as Patricia's go to guy or right hand man,
or something. I am interested if Patricia is. We would have to talk
but I think that a River implementation of JavaSpaces is important.
Is there any chance you could write a JavaSpaces benchmark? Perhaps work with
other pro-JavaSpaces people on the developer and user mailing lists to make it
realistic.
I'm worried that anything I do to fix the bug I'm hunting might slow it down.
In the long term, I want to make it faster.
As my next step, I'm planning to try to understand, and document, the
principles of operation of the current outrigger implementation.
Patricia
Michael McGrady
Chief Architect
Topia Technology, Inc.
Cel 1.253.720.3365
Work 1.253.572.9712 extension 2037
mmcgr...@topiatechnology.com
