Re: ZODB on pypy?

> On Oct 8, 2017, at 15:03, Etienne Robillard <[email protected]> wrote:
> 
> Hi all,
> 
> I would like your input on the feasibility of porting ZODB to pypy.
> 
> Is it possible to run ZODB on top of pypy?

ZODB already runs well on PyPy: https://github.com/zopefoundation/ZODB/blob/master/setup.py#L30 
As does ZEO: https://github.com/zopefoundation/ZEO/blob/master/setup.py#L47 
And RelStorage: https://github.com/zodb/relstorage/blob/master/setup.py#L31

Under PyPy, ZODB with FileStorage  produces numbers in the zodbshootout that are substantially faster than CPython (https://github.com/zopefoundation/ZODB/pull/160).(This is likely also true for ZEO with FileStorage, at least for sufficiently long-lived clients, but I haven't benchmarked recently.)

(Results in objects/second, higher is better.)
"Transaction",                    PyPy    CPython 2.7
"Add 1000 Objects",              35,651   19,531
"Update 1000 Objects",           54,906   16,201
"Read 1000 Cold Objects",       103,484   22,111
"Read 1000 Hot Objects",         84,721   21,851
"Read 1000 Steamin' Objects", 2,112,095  880,582

It's been a long time since I've run the benchmarks for RelStorage under PyPy (Predating the afor-linked PR160.). The most recent I've published are here: http://relstorage.readthedocs.io/en/latest/performance.html#pypy YMMV, especially depending on the database driver you use.

Keep in mind these are results of a synthetic benchmark designed primarily for tuning ZODB/persistent/RelStorage performance, so their correlation to actual application performance is unclear.

Jason

-- 
You received this message because you are subscribed to the Google Groups "zodb" group.
To unsubscribe from this group and stop receiving emails from it, send an email to [email protected].
For more options, visit https://groups.google.com/d/optout.