ClickHouse Cloud

I have experience in ClickHouse , and we also use Apache Druid , which has corporate support from Druid , along with data products in Hadoop . We are currently exploring many platforms such as GMI, TKI, and Vertex.

I use ClickHouse  as a merchant side portal, especially when we started exploring how to use the data, which was coming from multiple sources such as logs, mainframe, Teradata , and many file systems that come to the data lake. The real-time challenge was joining the data and providing more analytical queries for our merchants, who work throughout the year to improve GMB, sales, and ensure the right quantity of items is ordered at the right time. That's the challenge for the merchants, and we aim for fast analytical queries on larger databases, which is why we selected ClickHouse as our columnar OLAP database supporting real-time analytics with its own SQL interface.

We have installed both local Docker  versions, which are quite scalable, and usually connect with BI tools such as Grafana , Superset , and Tableau  while utilizing materialized views, DDLs partitions, and many other connectors with Python, such as ClickHouse connectors and drivers. It's exciting to see how ClickHouse has evolved, and we are evaluating ClickHouse Cloud while also having the on-premises version.

We are already a customer of ClickHouse, with Sam's Club utilizing it on the merchant side while also exploring ClickHouse for consumers, primarily for user analytics, metrics, and streaming data analysis in ad tech. Additionally, we use custom analysis and metrics for fraud detection in payments and ad campaign metrics, with various teams utilizing it for ad campaign management and user behavior analytics, particularly on e-commerce sites focusing on customer behavior. It's extensively used due to its low latency, fast aggregations, and excellent OLAP columnar storage, featuring quick joins and real-time data visibility, making ClickHouse very appealing to us.

ClickHouse is very easy to use; one of the good features is that it has joins, which were not present in Druid, and Druid was quite expensive, especially with our applications at Sam's Club utilizing ClickHouse very quickly.

ClickHouse deserves a rating of 9 when compared to competitors, particularly Druid, which is stable but comes with higher costs and subpar support. ClickHouse proves to be more lightweight, offering low latency and high throughput, along with joins, making it especially good for log and metrics handling.

The basic challenge for ClickHouse is the documentation, which isn't ideal, but it's mature and stable with more columnar storage, compression, and parallel processing, making it the best for OLAP. In terms of improvements, it's not designed for very frequent small writes, making it less scalable in write-intensive workloads, and it's not flourishing in transactional use cases or when ingesting streaming data, such as batching or buffering, which is something ClickHouse will improve.

ClickHouse is quite stable, and it deserves a rating of 9.

ClickHouse deserves a scalability rating of 8 since it's quite scalable but has some room for improvement regarding scaling challenges.

The support team has its own community support on platforms such as Slack  Overflow and ClickHouse Slack . Commercially, the company provides enterprise support, especially for Sam's Club through ClickHouse Cloud. We utilize AVN ClickHouse, which is effectively managed by AVN, providing bug fixes and developing new functionalities along with architecture reviews. I appreciate their 24/7 support which is beneficial, although those using open source might face some challenges. Overall, the enterprise support is quite good.

Positive

The initial setup for ClickHouse is relatively easier compared to Flink ; however, for newcomers, it is quite challenging. I find it easier in terms of API with single-node setups through Yum or apt taking only a couple of minutes to install. Planning cluster setups is a bit complex, primarily an admin task, and while a single-node setup is easy, managing ClickHouse Cloud is extremely easy. Creating clusters can vary from moderate to difficult based on the scale, typically from 5 to 10 nodes, depending on the use case.

I would recommend this solution. Overall rating: 9 out of 10.

On-premises

Other

The main use case for ClickHouse is as a data warehouse for handling large volumes of data that exceed the capabilities of traditional databases like Postgres. I use it for creating dashboards and performing analytical tasks such as determining the total number of orders, average order value, and evaluations and ratios for various stores. I deploy ClickHouse both on the cloud provided by ClickHouse itself and on-premises for IoT and similar data tasks.

One of the most valuable features of ClickHouse is that it is open source without vendor lock-in, allowing me the freedom to choose any vendor for the database. It offers numerous out-of-the-box analytical functions, eliminating the need for complex coding. The performance of ClickHouse aligns with its claims, being highly efficient and used by large organizations like Uber and Zomato. The deployment process is straightforward, and it is scalable both vertically and in distributed systems via the cloud.

A significant area for improvement is the documentation, which is not comprehensive and lacks centralized resources, making it difficult to find information. Additionally, ClickHouse lacks robust support for transactional data, which limits its use as a primary database. My developer experience could be enhanced through better-organized documentation, perhaps by offering a cheat sheet or centralized guide for common setup and usage scenarios.

I have known ClickHouse for more than two years, but I have used it for about one year.

Deployment is quite straightforward, though not all resources are directly on the official site. While it is not hard to find deployment information, having a cheat sheet on their site would be beneficial. Overall, I can figure out the deployment process within an hour or so.

ClickHouse is stable and performs exceptionally well with large data sets. It does not slow down under the volume of data that was problematic for Postgres.

ClickHouse is highly scalable. It is vertically scalable and can be used in distributed systems through their cloud service, managing scalability for large data volumes.

I have not directly contacted ClickHouse's support team but have joined their Slack channel where I asked a few questions.

Neutral

I previously used Postgres, which started slowing down with massive amounts of data. I evaluated over twelve databases, starting with TiDB, but found ClickHouse to be the best fit after considering options like DuckDB. I initially preferred Postgres for its comprehensive features, but it couldn't handle the data scale.

Initial setup is straightforward and not hard at all. I can figure out the process within an hour or so.

ClickHouse is open source without direct fees, unlike other databases that have hidden fees or restrict hosting to their platforms. The open-source nature of ClickHouse allows for complete flexibility without licensing constraints.

I evaluated over twelve databases, including TiDB and DuckDB, but I opted for ClickHouse based on its performance in benchmarks compared to others.

For the right use cases, I would rate ClickHouse eight to eight point five out of ten. However, it is not suitable as a primary database for startups due to the lack of transactional support. For companies with massive data struggling with query speed and facing high costs from vendor lock-ins, ClickHouse is an excellent choice.

Hybrid Cloud

Other

What do you like best about the product?

The best tool for quering millions of rows in one second, It is simple and easy and similar queries to sql but provides variation in queries. The integeration to logs is also awesome

What do you dislike about the product?

Nothing yes, it had made my work easy only

What problems is the product solving and how is that benefiting you?

It has helped me query millions of rows in no time that is seconds

Our company had about nine platforms, each with its own database and data. We had to gather all these data in one database and just one table. We used Apache Superset  to integrate this database with the business intelligence tool. We had too many choices or options initially for the database engine.

We initially tested a database, and its performance was good. When we tried ClickHouse , we switched to it immediately because the performance was really amazing. When we had a huge amount of data, about five or six gigabytes in just one table, and we used ClickHouse  to deduplicate some duplicated entries or records.

Clickhouse helped us to achieve our use cases with simple steps and good performance as mentioned previously

The main feature of ClickHouse is the optimizer, if we had too many records to deduplicate, the optimizer took this task by itself. The second valuable feature of the solution is its performance. It's not easy when we talk about five or six gigabytes of one table of data.

Then, if you have to generate too many KPIs, charts, lines, and reports, it's not easy to deal with all of these with just one engine and tool. ClickHouse was really nice in this respect, and we had no problem with its performance.

ClickHouse has its own concept of database triggers and doesn't support traditional database triggers.

11 months

We haven’t faced any stability issues with ClickHouse.

ClickHouse is a scalable solution.

I rate the solution’s scalability a nine out of ten.

We used the free, self-hosted community version of ClickHouse.

For about six gigabytes, we took about two seconds to fetch all data at the maximum performance. Otherwise, it was really nice to have a medium CPU or database engine and resources. We don't have a really huge server; it's just traditional servers and traditional resources.

ClickHouse is not a straightforward tool for anyone to use. Users need some time to switch from traditional things to study new concepts. 

We had just one client, Apache Superset . Apache Superset connects with just one connection but with too many requests. We had about 20 to 30 reports on the same page, and they work concurrently. 

The solution’s documentation is amazing.

I would recommend the solution to other users. ClickHouse is the first step to the next generation of databases. When we deal with this amount of data and this performance, I think it's a respected technology.

Overall, I rate the solution a nine out of ten.

Private Cloud

We use ClickHouse for a passive monitoring system in telecommunications. It is used to record primary data from the mobile network technology.

The tool is column-based and infinitely scalable. 

There are some areas where ClickHouse could improve. Specifically, we encountered incompatibilities with its SQL syntax when migrating queries from MySQL or SQL to ClickHouse. This difference in details made it challenging to figure out the exact issues. Additionally, we faced difficulties due to the lack of a proper Django driver for ClickHouse, unlike MySQL, which Django supports out of the box.

I have been working with the product for one and a half years. 

My company has ten product users. 

The company decided to use ClickHouse because mobile networks produce enormous amounts of data—millions of timestamped vectors, each representing a measurement, which total billions of rows per month. Initially, they used MySQL, but as the data volume grew, MySQL couldn't handle the load. Therefore, they switched to ClickHouse.

If you're considering using ClickHouse for the first time, my advice would depend on how much data you need to handle. For most scenarios where big data isn't involved, I don't think it's a good idea to use ClickHouse. SQL Server, MySQL, or PostgreSQL are well-documented and supported. The software you need to access these databases will be readily available. So, I don't see any reason to use ClickHouse for small to medium-scale scenarios.

I don't think you'll find it any more difficult than other databases, apart from the SQL syntax, which is a bit different. The most challenging part with ClickHouse is dealing with the large amounts of data it handles without overloading your server. I don't think the database itself is difficult to use. However, I was primarily accessing data from it and don't have much experience with setting it up or feeding it data. 

I rate the overall solution a nine out of ten. 

On-premises