As far as configuration, it depends on the cluster size and data sizes. It'd be nice to have the configuration defaults adapt based on the cluster size, but they don't today. I'm not too familiar with the YCSB tests.
FWIW, we initially planned to try to use YCSB internally for perf testing, but we ended up going a different route, creating our Pherf tool (http://phoenix.apache.org/pherf.html). The design goals were really different: Pherf is for baselining perf at scale over realistic workloads to get an idea of what perf looks like from release to release (it'll also do functional testing at scale). YCSB seems more targeted at comparing different back-end systems against each other. Thanks, James On Fri, Jun 12, 2015 at 3:24 PM, Andrew Purtell <[email protected]> wrote: > Thanks. > > That was more of a question for the Phoenix folks around here, if there are > any configuration options or particular JDBC idioms that might produce > better results than what the generic driver is doing.... > > On Fri, Jun 12, 2015 at 12:26 PM, Sean Busbey <[email protected]> wrote: > >> Yes, for now. I know that there are some problems using the JDBC driver >> with Oracle (see #128). >> >> I could put a note in our test plan to check Phoenix with the JDBC driver. >> It'd be even better if one of y'all could sign up to do the testing. We'll >> have a plan together monday. >> >> On Thu, Jun 11, 2015 at 12:10 AM, Andrew Purtell <[email protected]> >> wrote: >> >> > Pull #178 is only the existing JDBC driver with additional dependencies >> in >> > the POM, support files, and a different connect string, right? >> > >> > What more/different might it be than the existing JDBC driver? I'm >> thinking >> > of anything other than vanilla JDBC that Pherf might do. >> > >> > >> > On Tue, Jun 9, 2015 at 7:45 PM, Sean Busbey <[email protected]> wrote: >> > >> > > Hi folks! >> > > >> > > The YCSB community is reawakening, and we're prepping to start regular >> > > releases again. The first of these is hitting feature freeze on Monday >> > June >> > > 15th. >> > > >> > > There's an old PR for adding Phoenix support[1], but it has >> > unsurprisingly >> > > gone stale during YCSB's long hiatus. I'd like to get Phoenix support >> in >> > > sooner rather than later. We're aiming at monthly releases, so this >> > coming >> > > Monday isn't a hard deadline. >> > > >> > > Any chance someone from the Phoenix community could take a look? >> > > >> > > [1]: https://github.com/brianfrankcooper/YCSB/pull/178 >> > > >> > > -- >> > > Sean >> > > >> > >> > >> > >> > -- >> > Best regards, >> > >> > - Andy >> > >> > Problems worthy of attack prove their worth by hitting back. - Piet Hein >> > (via Tom White) >> > >> >> >> >> -- >> Sean >> > > > > -- > Best regards, > > - Andy > > Problems worthy of attack prove their worth by hitting back. - Piet Hein > (via Tom White)
