5 ms·

Not completely worthless: it does say something about the ease of getting good performance without having to know much about the ecosystem. But it doesn't spea

by fdr 14y ago

Not completely worthless: it does say something about the ease of getting good performance without having to know much about the ecosystem. But it doesn't speak to even the first "easy" 80% of what one would do to optimize performance, though. I appreciate this work with numbers, and the code to run it. I'm sure it can be adapted to take on these two suggestions.

bsg75 14y ago
> it does say something about the ease of getting good performance without having to know much about the ecosystem The root cause of most "I used [MongoDB,PostgreSQL,MySQL,CouchDB,...) and my site fell over" articles.
- lobster_johnson 14y ago
  "Getting good performance without having to know much about the ecosystem" is pointless, because choosing a tool is all about learning how to wield it. It's true about any tool, even something like a hammer. Software is complicated stuff and not something to treat superficially. Like everything in life, optimum results require a deep understanding. This article is worthless because the results are not useful to anyone for anything. Doing serious benchmarks requires a particular and fairly rare set of skills.
  - fdr 14y ago
    > "Getting good performance without having to know much about the ecosystem" is pointless, because choosing a tool is all about learning how to wield it. It's true about any tool, even something like a hammer. There is no complete understanding of any non-trivial system, even for its authors. There is a question of how much one can understand in a reasonable amount of time, and what fraction of a perceived ideal one can achieve. Some systems have steeper curves than others. > This article is worthless because the results are not useful to anyone for anything. Doing serious benchmarks requires a particular and fairly rare set of skills. It is not worthless because there is some code posted to run it. Someone could beef up the corpus (it's pretty small right now) and apply their tweaks (or try other data types like hstore, or other indexing strategies). This person tried his/her best and put something together. I agree the benchmark is pretty flawed, but given the thankless (and often adversarial!) nature of running benchmarks and publishing results, I'm grateful for anyone who does one and then publishes the code.

by fdr 14y ago

bsg75 14y ago

> it does say something about the ease of getting good performance without having to know much about the ecosystem The root cause of most "I used [MongoDB,PostgreSQL,MySQL,CouchDB,...) and my site fell over" articles.

lobster_johnson 14y ago
"Getting good performance without having to know much about the ecosystem" is pointless, because choosing a tool is all about learning how to wield it. It's true about any tool, even something like a hammer. Software is complicated stuff and not something to treat superficially. Like everything in life, optimum results require a deep understanding. This article is worthless because the results are not useful to anyone for anything. Doing serious benchmarks requires a particular and fairly rare set of skills.
- fdr 14y ago
  > "Getting good performance without having to know much about the ecosystem" is pointless, because choosing a tool is all about learning how to wield it. It's true about any tool, even something like a hammer. There is no complete understanding of any non-trivial system, even for its authors. There is a question of how much one can understand in a reasonable amount of time, and what fraction of a perceived ideal one can achieve. Some systems have steeper curves than others. > This article is worthless because the results are not useful to anyone for anything. Doing serious benchmarks requires a particular and fairly rare set of skills. It is not worthless because there is some code posted to run it. Someone could beef up the corpus (it's pretty small right now) and apply their tweaks (or try other data types like hstore, or other indexing strategies). This person tried his/her best and put something together. I agree the benchmark is pretty flawed, but given the thankless (and often adversarial!) nature of running benchmarks and publishing results, I'm grateful for anyone who does one and then publishes the code.