Sisense

amazing>>>

Builds required everything to be in-memory and took too long. A system like Snowflake, Databricks, BigQuery or Redshift (RA3) that separated out storage from compute allows for a continual stream of data to be processed and queried against. The builds took hours and hours, sometimes we could only do 2 a day. It felt like the 90's with OLAP and not a newer columnar database. Partial builds, caches and lots of other hacks exist but ultimately the latest generation of data warehouses that separate storage from compute allows for faster (and cheaper) handling of costs. As queries increase, compute nodes increase. Additionally we experienced lockups and botched upgrades. It's big installs, bad patches, etc. If you don't have terabytes or more of data, it'll probably work but it just didn't scale. For smaller sites that need an all-in-one solution with decent front-end, simple connectivity in the back and don't mind the builds then it'll probably work. It worked for us during the demo and in the first 3 months of implementation but then every small thing seemed to turn into a big issue and compounded upon itself as it grew. Tableau on Snowflake or some similar setup is probably getting the best of both worlds. Move data around with Kafka, Nifi/Airflow or pickup another vendor to fill in that gap if your needs are more complex.