Apache Druid vs. ClickHouse

It is extremely well suited to rapid ingest of data from large data sources, due to the fact that you can restrict what is ingested by column/field, so that you only pull in the data you actually want or need.

As stated earlier, the open source version could use better cluster management tools, and troubleshooting tools for failing jobs/tasks.

The most important thing when using ClickHouse is to be clear that the scenarios in which you want to use it really are the right ones. Many users think that when a database is very fast for a specific use case, it can be extrapolated to other contexts (most of the time different) in which a previous analysis has not been carried out.

ClickHouse is an analytical database, as such, it should be used for such purposes, where the information is stored correctly, the data volumes are really large and the queries to be performed are not the typical traditional queries on several columns with multiple aggregations. ClickHouse is not the solution for this.

On the other hand, if your case is not one of the above, it is quite possible that ClickHouse can help you. Where ClickHouse shines is when you are looking for aggregation over a particular column in large volumes of data.

Incentivized

• Rapid ingest
• Limiting ingest to only the relevant fields/columns
• Easy ingest spec creation

• Their MergeTree table engine provide impressive performance for data insert in bulk
• Not only data insert but also the way MergeTree engine uses Primary Keys to sort the data and perform data skipping based on the granules its also their secret for ridiculous fast queries
• Data compression its also great
• They provide especial table engines that allow you to read data directly from other sources like S3
• Since its written with C++ you have very granular data types and especial ones like enum, LowCardinality and etc, they save you a lot of storage since are stored as integer values
• ClickHouse functions besides the ones that respect ANSI Standards are also awesome and useful

Incentivized

• Security configuration is problematic
• Cluster management could have more features
• Troubleshooting incomplete tasks/jobs is a chore

• Avro data manipulation
• Kafka consistency
• DDL operations errors (by replica configuration)

Incentivized

ClickHouse outperforms, especially in costs, since its compression/indexing engines are so smart, and even with very low computing power, you can already perform huge analyses of the data.

Incentivized

• Integration with S3 storage has saved about 35% on our storage, over HDFS
• The rapid ingest has saved user's time in the query aspects of their applications.
• The ability to ingest from a variety of data sources has made overall user application queries much simpler

• Queries that used to take more than 2 minutes now take less than 1 second
• Possibility to analyze use cases in real time (before was impossible)
• The applications are more complete and the users decisions are better

Incentivized