5 ms·

I will shamelessly plug myself again. I am the main author of RediSQL [1] and I am about to launch a managed service for it. It will allow to write SQL (SQLit

by siscia 7y ago

I will shamelessly plug myself again. I am the main author of RediSQL [1] and I am about to launch a managed service for it. It will allow to write SQL (SQLite dialect) against an HTTP or Redis protocol, to make thing clearer: https://simplesql.carrd.co/ https://simplesql.carrd.co/ Eventually it will upload your database to an S3 bucket for backup. [1]: https://redisql.com/ https://redisql.com/

erezsh 7y ago
Looks pretty cool! I can imagine it would be useful for data-science projects, if the pricing is right.
- siscia 7y ago
  What pricing would feel right to you?
  - erezsh 7y ago
    I'm far from an expert on such things, but I like the adjustable fixed-price model (like digital ocean). Whenever I see a price like 0.0002$ per hour, I feel like they're trying to mess with my intuition. That's especially applicable to data-science, because you're not worried so much about automatic scaling, you just don't want to be surprised at the end of the month. I don't know if you can match digital ocean's prices (they offer managed postgres, for a fair comparison), but if you can get close then you have a chance.
    - siscia 7y ago
      I was thinking something based on size or performance. 1 req for second, free. 50 req for second, 5€. Unmetered, 30€. Maybe not counting for seconds but for hour or day so to accommodate burst. Another option would be on features. But that will requires time.
      - erezsh 7y ago
        If you have full support for SQL, it means some queries can run for a few seconds, possibly even minutes (even sqlite supports some form of bfs). So will you just time-out those requests? That would make a lot of possible uses suddenly impossible. I might consider the unmetered option, if the performance gain is worth the 6x cost of other providers. For web and games, maybe your existing model could work, but then always having a recent backup also becomes more important (and for most sites, the lack of acid is a deal-breaker). The nice thing about data-science is that you interact with the workspace yourself, so you know exactly when you want to save a snapshot.
        siscia 7y ago
        Hummm, what you mean by data science? Like running analysis in a Jupiter notebook? Indeed I am not sure this case is a good fit...
        erezsh 7y ago
        Yes, that's what I meant. And possibly for ML preprocessing. Just out of curiosity, who do you imagine your users will be?
        siscia 7y ago
        The problem with data science is that usually you have relatively big datasets, you care more about throughput than latency and you work in secure an environment where you can definitely have access to the database credentials. Streaming over the network the result of a big select is not ideal, moreover I believe that data scientist prefer to work with common technologies. I mean that there are already adapter for SQLite or PG or MySQL, while for RediSQL it won't be as straightforward. I am thinking to developers for the JAM stack would be interested to this sort of API. Or people that want a database without having to think too much about it.
        erezsh 7y ago

by siscia 7y ago

erezsh 7y ago

Looks pretty cool! I can imagine it would be useful for data-science projects, if the pricing is right.

siscia 7y ago
What pricing would feel right to you?
- erezsh 7y ago
  I'm far from an expert on such things, but I like the adjustable fixed-price model (like digital ocean). Whenever I see a price like 0.0002$ per hour, I feel like they're trying to mess with my intuition. That's especially applicable to data-science, because you're not worried so much about automatic scaling, you just don't want to be surprised at the end of the month. I don't know if you can match digital ocean's prices (they offer managed postgres, for a fair comparison), but if you can get close then you have a chance.
  - siscia 7y ago
    I was thinking something based on size or performance. 1 req for second, free. 50 req for second, 5€. Unmetered, 30€. Maybe not counting for seconds but for hour or day so to accommodate burst. Another option would be on features. But that will requires time.
    - erezsh 7y ago
      If you have full support for SQL, it means some queries can run for a few seconds, possibly even minutes (even sqlite supports some form of bfs). So will you just time-out those requests? That would make a lot of possible uses suddenly impossible. I might consider the unmetered option, if the performance gain is worth the 6x cost of other providers. For web and games, maybe your existing model could work, but then always having a recent backup also becomes more important (and for most sites, the lack of acid is a deal-breaker). The nice thing about data-science is that you interact with the workspace yourself, so you know exactly when you want to save a snapshot.
      - siscia 7y ago
        Hummm, what you mean by data science? Like running analysis in a Jupiter notebook? Indeed I am not sure this case is a good fit...
        erezsh 7y ago
        Yes, that's what I meant. And possibly for ML preprocessing. Just out of curiosity, who do you imagine your users will be?
        siscia 7y ago
        The problem with data science is that usually you have relatively big datasets, you care more about throughput than latency and you work in secure an environment where you can definitely have access to the database credentials. Streaming over the network the result of a big select is not ideal, moreover I believe that data scientist prefer to work with common technologies. I mean that there are already adapter for SQLite or PG or MySQL, while for RediSQL it won't be as straightforward. I am thinking to developers for the JAM stack would be interested to this sort of API. Or people that want a database without having to think too much about it.
        erezsh 7y ago