A Question of Speed

[Gavin_Attard]

Hi @knime

So i saw this post on Linkedin and wondered on thoughts with regards to comparison and perhaps areas of potential improvement.

[Matt_D]

Would like to know if AMP was used in Alteryx or KNIME Columnar Table Backend in KNIME. Although not out of the box features it does suggest both pieces of software are pushing for more speed.

At the end of the day, were talking seconds not minutes or hours and although yes speed in Alteryx is something to be smug about sometimes it lacks some flexibility and configuration, parsing JSON being a great example. As long as the job gets done right and is easy to understand and explain to others I’ve become unfazed by the speed novelty. I’m not fussed anyway

[Gavin_Attard]

I lean on your thinking, 2 or 10 sec is not going to phase me or have a massive impact in the grand scheme of things. though i suppose in this crazy world of instant it’s always something to look at improving

[Luca_Italy]

indeed, i’ve made some test with an older version of alteryx (11.3 maybe) and the new columnar table backend: knime is on par, in some cases just better than alteryx.

[wiswedel]

Thanks for sharing!
In the platform design we have always chosen flexibility over performance but there were several additions in the past releases: the columnar table backend (not used here?), the re-write of the file reader (not used here?), streaming vs node-by-node execution (also not used here?). Especially the latter could be an interesting alternative as the workflow filters data right after the reader. AFAIK Alteryx does only streaming. Anyway, we’ll continue to invest in performance improvements also in the future.
– Bernd

[Gavin_Attard]

@wiswedel - Thanks for the insight, and would support the approach. Interestingly the same user posted another test, this time comparing ML processes and Knime just wiped the floor…

You also brought up some interesting configuration options which should be clarified, perhaps you or your colleagues are best placed to put them to the user, it’s Bob Peers (Invisio Consultant)