Our schemaless data ingest + automatic indexing definitely draws a lot of inspiration from search based systems such as ES and Solr. And yes, the biggest difference here is that Rockset allows
1/ full featured SQL (with fast JOINS, aggregations, sorts etc) on such semi-structured data sets, and that
2/ it is built ground up to exploit cloud economics and scale (which is why we are able to offer this as a serverless data management system).
Our goal is to make the default experience simple so that you get enough compute to build most real-world apps and dashboards. We still give you flexibility in case you want to purchase additional compute for ingest or queries. We will make this clear in our pricing page -- thanks for the feedback.
> - What do I do if I need more QPS?
Barring extreme workloads (say 1 million QPS on 1 GB of data) for which we are not a good fit anyway, we auto-scale enough compute to handle the QPS needs of most real-world applications. As I mentioned earlier, if you want to break out of the standard compute allocation, then we do offer ability to purchase additional compute, but from our experience, this is seldom required.
> - What do I do if I need to ingest more data.
Yes, you can purchase additional ingest bandwidth if you need a higher steady state ingest capacity. Please note that the bandwidth limit only applies to real-time streaming ingest — for bulk ingest (for example: the first time a collection is created in Rockset sourced from Amazon S3) we try to build the indexes at much higher speeds and will keep working on making that really really fast without any additional fees.
Yeah, we are fans of Metabase too and we will soon add support for connecting Metabase with Rockset. We do have Redash [1] and Superset [2], which are also pretty good and open source.
Good feedback. We thought about the different ways to frame the value prop and "serverless" is what resonated the most with us because:
1/ you can load data, process queries and build apps/dashboards without ever thinking about servers -- so, no provisioning or capacity planning required.
2/ you only pay for amount of data actually loaded and indexed -- so, no idle servers costing you $$$s.
If you have better suggestions that feels more accurate, please share and we will definitely consider it.
Touche on the "cloud hardware" bit. We will fix that soon.
Your assessment here is spot on @itronitron
Our schemaless data ingest + automatic indexing definitely draws a lot of inspiration from search based systems such as ES and Solr. And yes, the biggest difference here is that Rockset allows
1/ full featured SQL (with fast JOINS, aggregations, sorts etc) on such semi-structured data sets, and that
2/ it is built ground up to exploit cloud economics and scale (which is why we are able to offer this as a serverless data management system).