[Date Prev][Date Next]
Re: benchmarks &real world
"Quanah Gibson-Mount" wrote :
> And it also depends on the number of attributes in a given entry. 10,000
> entries with a single attribute are a lot more lightweight than 10,000
> entries with 60+ attributes.
And even the length of the entry/attribute makes a difference, I believe -
correct me if im wrong here - as the performance of the ASN.1
encoding/decoding may make a significant difference.
Which probably means that it will be all that much more easy to 'compare' to
other systems/vendors/products, if you use a somewhat 'standardized' test,
like for example MindCraft's DirectoryMark. This test always uses the same
amount of entry's, the same attributes, attributes length etc. etc and then
reports this as 'Operations / per second'. Which still leaves the question
of CPU/MEM/DISK etc, which is why it also reports a price/performance
metric, which is expressed as the 'System Under Stress' price in dollars,
divided by the 'operations per second'. So that actually makes it relatively
fair and/or easy to compare different kinds of product/systems.