On 21/08/14 06:43, Deyan Chen wrote:
Hi Andy,
I see the performance test results of SDB on different SQL DBs in Jena
website, but I don't find similar tests for TDB.
I want to know that if there are some performance evaluation results for
TDB.
Thank you very much.
Deyan Chen
Hi,
The SDB results compare the two SDB choices (hash and index) and are
getting to be rather out of date. TDB does not have the same
possibility of different database layouts so I don't know what we'd publish.
There are BSBM runs of TDB so they are independent evaluations.
LUBM isn't a very helpful benchmark and anyway the "right" benchmark is
what you are going to use any system for. The style and nature of the
queries matetr, as does the hardware setup. And these days, hardware,
especially the amount of available RAM, is getting better at a
significantly fast rate.
If I were doing a new SPARQL database [*], I'd design more for in-memory
with disk/ssd as a way to retain state across reboots ad crashes.
There is the "Linked Data Benchmark Council" putting this on a footing
like TPC. It is a pay-to-play organisation - a system needs to pay to
have the tests run in and also provide and tune the hardware. It skews
things towards commercial systems who can afford that.
Andy
[*] I am - but it's not a single machine database.
