perfsonar-dev - LS Benchmarking (again)
Subject: perfsonar development work
List archive
- From: Maciej Glowiak <>
- To: Jeff Boote <>, Jason Zurawski <>, Szymon Trocha <>, Martin Swany <>
- Cc: Perfsonar Development List <>, Roman Lapacz <>, Ben Perry <>
- Subject: LS Benchmarking (again)
- Date: Fri, 27 Apr 2007 12:45:28 +0200
- Face: iVBORw0KGgoAAAANSUhEUgAAADAAAAAwAQMAAABtzGvEAAAABlBMVEUAAAD///+l2Z/dAAAA CXBIWXMAAEU1AABFNQF8gVf5AAAAB3RJTUUH1QYQDjo6uEWvwgAAAM5JREFUGNNN0LFqAkEUheGj KRZsfATrvENgYyH4APabxwgWGUUQC99BsNDCInUq7VImbbDZ0kayxBXMuN7jvTuKVh//mZlmQKZ1 EhQ8GAVgZECspEBdWQHRjR70KlgFKkoUaCw3ijSYQ4n5HfBK4a4jDcdDQPol/80Sr9BxZOOL4Fmr Jq8VBx7eopaSPvWGOm67fqol3j1q0XNs7Nk2cs6MU6gPNzf+ZGKQX4Ek8H6rAnFZnXB2vJxJcv8g C2P+WzL4tD+Txc4KydrIkh+eAdo01QbjQ84vAAAAAElFTkSuQmCC
- Organization: Poznan Supercomputing and Networking Center
Hi guys,
This week I finished new way of accessing XML DB by Lookup Service. As you probably saw, I implemented it in two ways. One is using XQuery modifications (instead recent XUpdate), the second one using "Multiple files" where each Lookup Information is stored inside different XML file.
In the first option not only XUpdate were converted into XQuery Modifications but I moved a lot of code into XQuery expressions, so I optimized the way how it worked. Now more "logic" of LS was put into XQuery expressions what seems to be more efficient.
The second option wasn't fully implemented (I still need to work on LS Cleanup, but it won't be so time consuming I hope), but all basic functionality works fine. While registration Lookup Information goes to separate XML file inside Exist DB XML collection. Parameters are also stored separately. Because of some limitations of eXist DB XML or/and HTTP requests I needed to use MD5 encoding of file names, but it's an implementation detail and is transparent for user querying LS or service storing information (that's because XML file names should be in fact URLs and eXist DB XML doesn't support it through HTTP).
I also updated our benchmarking code a bit and re-run the procedure again, testing both options.
Results are in the attachments. All times are in seconds. Date used in both tests were exactly the same. Database was cleaned up and then they both with LS were restarted.
It seems the second option is much more efficient, but still needs some development (LSCleanup) and testing (which will be done in next weeks - the next week, when some of you are in Brazil, we have holidays in Poland).
I just wanted to give you some information for Salvador Meeting discussions.
Best regards
Maciej
--
--------------------------------------------------------------------
| Maciej Glowiak Network Research and Development ||
|
Poznan Supercomputing and Networking Center ||
| (+48 61) 858 2024 -- skype_id: maciej_psnc GG: 4526858 ||
====================================================================
Attachment:
results-test1.pdf
Description: Adobe PDF document
Attachment:
results-test1.xls
Description: application/msexcel
- LS Benchmarking (again), Maciej Glowiak, 04/27/2007
- Re: [pS-dev] LS Benchmarking (again), Szymon Trocha, 04/27/2007
Archive powered by MHonArc 2.6.16.