edlich/nosql-database.org

performance-based list ordering?

Opened this issue · 3 comments

A comment in #54 said --

What is the performance? (quite brave to put yourself before Mark Logic)

This is the first suggestion I've seen that the list ordering is based on anything other than randomness.

As you now seem to say that "performance" (a rather opaque and variably defined word) is supposed to guide the order of these listings, it seems reasonable that some notes should be added to the page -- minimally including what benchmark(s) are being used to determine this comparison, and how new entrants may produce verifiable results to guide their own placement on the list. Minimal information required would include:

  • what hardware (or cloud VM setup) should be used?
  • what network or other tuning should be applied?
  • what dataset should be used (if not specified within the benchmark)?

LDBC and TPC benchmarks are reasonably well specified along those lines, though it is entirely possible to run any benchmark against a poorly configured instance and get terrible results, where a properly configured instance would deliver great results.

I would especially like to know these answers so that Virtuoso can be put in its proper positions in all sections where it is (or should be) listed.

Hi Ted,
the list is not ordered by performance benchmarks.
In fact it is not really ordered. Perhaps a little by popularity.
If you feel something should be rearranged, please tell me so.
best
sed

I would suggest some ordering, by something a bit less ephemeral and subjective than "popularity".

The most common (dare I say, "default") ordering for such lists is alphabetical by product or, less often, vendor name. Such ordering is typically visually obvious, letting visitors easily locate and/or appropriately add items of interest. I would be happy to produce a PR for such, if you would apply it.

Hi Ted,

  1. thanks for your Energy. But I do not want to make it alphabetically. In this case some weird DBs might appear above which I do not like.

  2. I am not fully lucky with the way OpenLink Virtuoso been presented now:

  • The format is different. There are many CR/LineFeeds in there. No database description has this.
  • One database should be in there only once. E.g. Arango DB is in several categories, but mostly with a link to multimodel. It should be like this too for OpenLink.
  • In the feedback section I speak about a 500 char limit. Every vendor somehow likes to break it ;-)

If you do not sent a PR I would change it for your product.
Best
Stefan Edlich