Ciao Luigi,
so, in your case, you want to compare two different software systems
on the same hardware. If it is not on the same hardware, it becomes
more difficult to make meaningful comparisons.
I suggest you look at the BSBM documentation, get the code, generate
your datasets and run the benchmark against SDB+Oracle vs. TDB.
Even if you do not use bsbm-automated, you can look at the Bash scripts
there to quickly learn how to download/compile/run BSBM, how to generate
datasets. There is also an sdb.sh (it does not use Oracle, but it should
not be too difficult to adapt to it).
I do not think there is a mailing list for BSBM questions, if there were
one you could/should ask directly to them why Jena SDB has been removed
from the BSBM after 2009. I do not know (but I could guess) why.
Go ahead and let us know your findings.
Paolo
Luigi Selmi wrote:
Ciao Paolo,
thanks for the advice and links. My purpose is to compare Jena SDB with Jena TDB but using Oracle as the DBMS instead of MySQL since I'm working in a project in which we have been requested to use the DBMS that is available in test and production environments in case we'll use Jena SDB. Btw Jena SDB has been removed from the BSBM after 2009.
Luigi
Date: Thu, 3 Nov 2011 13:47:38 +0000
From: [email protected]
To: [email protected]
Subject: Re: Jena SDB + Oracle Benchmark
Ciao Luigi,
the best benchmark data you can find is the one you produce yourself.
Benchmark data has usually little meaning in absolute terms, it's more
useful to make relative comparisons for the same software, same
configuration against different hardware. Or, same hardware to compare
different software systems. Or, same hardware, same software and
different configurations/settings.
My best advice to you is: learn how to use/install/configure SDB.
Same for Oracle 10g (and tuning the performances). Install the software
on your hardware.
Learn how to use BSBM and run your own experiments and produce your
numbers.
Also, this way you have everything ready to run benchmarks against
your own data, using SPARQL queries derived from your own use cases.
Unfortunately, no SPARQL benchmark is currently very easy to download
and run (out-of-the-box). This is IMHO an issue, because a benchmark
should be really easy to run:
1. download {filename}.zip
2. unzip
3. run(.sh|.bat) http://{hostname}:{port}/{sparql end-point}
Anything more than this is a too high barrier for most of the users.
IMHO, the value of a benchmark is not in the benchmark itself, but
the fact that it is used by a lot of people and therefore more results
become available and it becomes easier to compare different systems/
settings. It should be extremely easy for an user to run a benchmark
and produce a "standard" report that if you want you can upload
somewhere.
IMHO none of the current existing SPARQL benchmarks achieved this in
terms of usability and wide spread adoption.
Since you specifically referred to BSBM...
Not an answer to your question or your needs and not something which
is trivial to download and run, but something I found useful is here:
https://github.com/castagna/bsbm-automated
See also:
- https://github.com/afs/BSBM-Local
- https://svn.apache.org/repos/asf/incubator/jena/Experimental/JenaPerf/
- https://github.com/castagna/sp2b-automated (this is for SP2B)
Paolo
Luigi Selmi wrote:
Hi,
I am looking for some benchmark data about the performances of Jena SDB and Oracle 10g for load and response time. The only data I am aware of is the Berlin SPARQL Benchmark that is quite old (2009) and based on MySQL 5.1 for storage.
Thanks in advance
Luigi
