ClickHouse Reviews

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.

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

My main use case for ClickHouse is data ingestion and for its OLAP properties, as we had use cases where database locks were slowing us down and because ClickHouse does not have that, we chose to use it.

I could give you a quick, specific example of how I'm using ClickHouse for data ingestion where the lack of database locks helped us when we were parsing IOCs among other things, as a lot of that data has to be processed really quickly and ingested into the database for further processing and identifying which IOCs are compromised.

ClickHouse helped us solve the problem that we were having and it's one of the two databases we used at Cyware—one was Postgres, the other was ClickHouse.

The best features ClickHouse offers are its OLAP features because, given that there are no database locks and its eventual consistency, that is the biggest feature that we have or that solved our problems.

The eventual consistency and lack of database locks specifically benefit my team in terms of speed and reliability, as once data is ingested, we have to quickly process it and then show the outputs to the user, say there are ten indicators of compromise, and we have our own database where we tally whether these IPs or IOCs that we are scanning right now are marked or red flagged before or not, so we have to quickly scan them, process them and then give an output, and that helped us with the reliability part, the speed part, while eventual consistency is used on a different side of the product.

ClickHouse has positively impacted my organization, as there was an entire exercise done on which database we were supposed to use for solving our problems, and we found ClickHouse was the one performing the best, which is when we adopted it.

ClickHouse can be improved on the documentation side, and there is one small constraint that is mentioned in ClickHouse documentation, which is a partition limit of ten thousand that we hit, so if that can be increased or there are workarounds around it, that would be great.

I chose nine out of ten because, as I mentioned, the improvement side and the ten thousand partition limit created issues that we were hitting quite frequently, but with some schema manipulations we did manage to find a workaround, although that could have been avoided had things been better documented on how we could have solved this problem in a different approach, which took some bandwidth.

I do not have any other improvements I think ClickHouse needs, besides the documentation and partition limit.

I have been using ClickHouse for about a year, maybe slightly more than that.

ClickHouse is stable, as we did not encounter stability issues in production, but in the dev environment, one of the seniors did flag one specific point where we found some inconsistencies, although I think they did find a workaround around it, but it was stable for us.

ClickHouse's scalability is good.

We previously used Postgres, and we encountered issues with Postgres, which was again, as I mentioned, why we did a study on switching.

I have seen a return on investment, as I can share that on the engineering side we had improvements in database performance, but for the metrics asked, time saved, fewer employees needed, I do not have them.

My experience with pricing, setup cost, and licensing was such that the setup costs were just my own bandwidth, while licensing and pricing were done by other members of the team so it was abstracted away from me, and I am not aware of it.

Before choosing ClickHouse, we evaluated other options such as Apache Druid and Pinot from Apache, and then there was a study.

The advice I would give to others looking into using ClickHouse is that on the engineering side, if there is some OLAP use case or anywhere where data needs to be ingested at very high rates or there is a use case for eventual consistency, then perhaps it can be used. I gave this review a rating of nine out of ten.

Hybrid Cloud

Amazon Web Services (AWS)

I use ClickHouse to collect and analyze data from Ethereum. We primarily use it for data classification and occasionally for machine learning with GPT, but that's minimal. The primary use case is classification; sometimes, we use it for applications similar to OLTP scenarios. All of our data is stored in ClickHouse. We are customers of ClickHouse, not partners. It's an easy tool to use if you know SQL databases.

The aggregation capability is a valuable feature. It's highly efficient, allowing us to review entire transaction histories and user activities in the market. We've tried MongoDB, Postgres, MariaDB, and BigQuery, but ClickHouse is the most cost-efficient solution for collecting data at high speeds with minimal cost. We even used ClickHouse Cloud for a month, and it proved to be a great setup, especially for startups looking to handle big data. For example, if there is a need for 2-4 terabytes of data and around 40 billion rows with reasonable computing speed and latency, ClickHouse is ideal.

Regarding the real-time query performance of ClickHouse, when using an API server to query it, I achieved query results in less than twenty milliseconds in some of my experiments with one billion rows. However, it depends on the scenario since ClickHouse has limitations in handling mutations. 

Additionally, one of ClickHouse's strengths is its compression capability. Our experimental server has only four terabytes, and ClickHouse effectively compresses data, allowing us to store large amounts of data at high speed. This compression efficiency is a significant advantage of using ClickHouse.

We faced a challenge with deploying ClickHouse onto Kubernetes. Recently, we've been using ClickHouse Cloud, and the main issue is the high cost of the cloud service. The pricing isn't very competitive, especially for startups. I would instead buy a server and self-host if I have enough disk space. Besides that, ClickHouse has done very well, with clear goals and effective execution.

I have been using Clickhouse for the past one and a half years.

Based on stability, I would rate ClickHouse around nine out of ten.

The cloud services support is excellent. Their support team is very timely and helpful, and even if you encounter any bugs, they assist you quickly. Compared to other services I've used, ClickHouse's support is very helpful. Even if you don't know much about databases or ClickHouse, their support will help resolve any issues.

Positive

For pricing, if you use the self-hosted version, it would be free. Cloud services pricing would be an eight out of ten. I try to minimize costs but still have to monitor usage.

 I would recommend ClickHouse to others. If they have large datasets, ClickHouse is much more cost-effective and efficient than BigQuery. For example, running a query on one billion rows in BigQuery took a few minutes and was very expensive, whereas ClickHouse could do it in less than five seconds at a much lower cost.

I don't use AI in ClickHouse, but I use full-text search, and it's mighty. There's no significant gap when migrating from other SQL databases to ClickHouse, though you must learn some specific syntax. If you are familiar with databases and know how to code and design systems, using ClickHouse should be straightforward.

Overall, I would rate ClickHouse an eight out of ten.

I used ClickHouse to collect data, put it in the database, and then analyze it to find insights. The main advantage is that I can install it on my computer instead of using cloud-based solutions, so I don't have to pay for every query like with Google or Amazon cloud services.

The best thing about the tool is that I can set it up on my computer and run queries without depending on the cloud. This is why I use it every day.

There aren't too many improvements I'd suggest for ClickHouse as it covers all my needs. There are just a few technical issues. For example, sometimes, when you want to get unique values and use certain tables, they don't work as expected. But it's not a major problem.

I have been using the product for two years. 

The solution is stable and works well. However, it consumes a lot of memory, so you need plenty of RAM in your computer or cloud solution to run it effectively. This could be a problem because you need to think about how much memory you have for calculations.

In my company, many people used ClickHouse—over 1000 people could access it. About 10-20 people used it at a professional level, creating tables and maintaining the database.

I have talked to the ClickHouse support team before. They have a support group on Telegram messenger where you can ask technical questions. I often asked about working with tables and views and making sophisticated calculations. But now, I don't have any issues, so I don't need to ask for support.

I was satisfied with the support. Many people in the support group try to really understand your problem and help, not just dismiss it. If something isn't possible due to database limitations, they try to help you look at the situation differently.

Installing the tool was easy. I used a Windows laptop with WSL and followed the documentation instructions. I didn't have any issues with the installation.

The tool is open-source. 

We chose ClickHouse because we needed to move away from cloud-based solutions due to risks in Russia. We considered other options, such as Postgres, NoSQL databases, Hadoop, and Hive, before deciding on ClickHouse.

The tool is open-source, so you don't need to pay for the software itself. However, you need to consider hardware costs and maintenance. A small company can install it on a company computer. For larger companies, you might need to hire a team for maintenance and consider data safety and privacy issues.

Integrating ClickHouse with other tools in our data stack was easy. It has native connections to many tools, such as Google and Amazon cloud solutions, and can easily connect with other databases.

For beginners, the ease of use depends on your background. If you're familiar with relational databases, it's easy. If not, you might need to read the documentation or ask for support.

On-premises

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 the solution not just as an analytics database but also as a data warehouse. A lot of our internal services communicate with the tool. Our analytics and ML teams also use the solution. It's the hub of our company.

It's easier to work with big data and calculations using the product. For example, we can easily calculate metrics around terabytes of data using ClickHouse. The dictionaries help us to make our analysts’ jobs faster and easier and give value to the business faster.

The clusters are not perfect. We had a lot of troubles while deploying a whole cluster. We must tune some sequences, so we must have experience with the product. I worked a lot with bare metal. However, working with the cloud is a little bit harder. When we need to start up and shut down some nodes, we need to start or shut down the whole cluster. It is not so in Databricks.

I have been using the solution for almost eight years.

There were a lot of bugs before. Now, it's less. Open-source tools contain bugs. Any technology created by humans will contain bugs. The bugs are critical sometimes, but they are always updated. I rate the stability an eight out of ten.

The tool is scalable. I rate the scalability a ten out of ten. We have 20 users in our organization.

The product provides a lot of community support. It is useful. We can also contact a private company that provides support for ClickHouse. The solution has a community chat in Telegram that works well. We find solutions easily when the problem is already mentioned by someone. In rare cases, the issues stay unresolved because of NDA.

Neutral

The initial setup is easy for me because I have a lot of experience. Generally, when I mentor someone to deploy the whole cluster, it is difficult for them. It will not be hard to deploy one server or node because ClickHouse’s team has shown a great demo of deploying it as a service on Linux. It gets harder if we want to tune some small sequences to get more performance. Real-time models require experience. We can open some community chats and find help. The deployment can be done in a couple of minutes. It’s very fast.

The tool is free. It is open-sourced. However, if we do not know how to deploy it and are unwilling to support everything, we can contact the vendor and create a cloud version. It is cheap, but it depends on the scale.

We do not use the real-time features much. Usually, we work with big data. We do not need to work with big data in real-time. We use CatBoost with ClickHouse. I always recommend the tool to others based on their requirements. If you have trouble with your queries and think that ClickHouse is slow, please review your queries. Overall, I rate the solution a ten out of ten.

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

We use it to store live data and for tracking and monitoring every action on a PC, like which websites are opened, through Kubernetes and Google Chrome. Data is sent every second from ActiveMQ, and ClickHouse can insert millions of data points per millisecond.

ClickHouse is much faster than traditional databases like MySQL and MongoDB. Its column-row searching strategy makes it very efficient. With ClickHouse, we can manage multiple databases, automatically insert data from other databases and delete data as needed. It supports real-time query performance, allowing simultaneous data insertion and retrieval. ClickHouse has improved significantly over the past two years, adding more functions and queries, as well as top functionality.

Initially, I faced challenges integrating ClickHouse, particularly with inserting data from ActiveMQ, which caused duplicates. However, after adjusting the ClickHouse settings, the issue was resolved and there were no more duplicates.

I've been using ClickHouse for three years in a product-based company.

I have not faced any stability issues and I would rate it a seven out of ten. 

It's a simple database, similar to MySQL, and can also be used as a NoSQL database. While ClickHouse provides most of the essential functions, learning and understanding some of its methods can be difficult.

ClickHouse is very scalable, and many companies, including Uber, are using it. For scalability, I would rate it an eight out of ten.

The ClickHouse support team is small but useful. I haven't needed to use their support often because I usually find solutions through ClickHouse Docs.

ClickHouse is easy to use, with a straightforward initial setup.

The return on investment is high and gives value for money.

ClickHouse Cloud is not expensive compared to other databases, costing a few dollars per month while providing fast performance. 

Although I haven't used AI with ClickHouse extensively, it's a great option because ClickHouse can handle large data volumes and perform queries very quickly.

I would recommend ClickHouse to others, especially for real-time applications like chat, map locations, and AI tools. Compared to MySQL, ClickHouse handles large datasets and queries very quickly, making it a perfect choice.

Overall, I would rate ClickHouse a ten out of ten.

Our use cases are for data analytics, both real-time and batch, and also for logging Clickstream data.

We use it in our organization. We have it in our production environment.

The query engine is super fast. We deploy ClickHouse on our Kubernetes cluster, not as a cloud subscription, so it's easy to scale with the deployment.

Some features, like connecting to third-party applications or the cloud, could be better.

I have been using it for one year. 

One issue is that you need persistent volumes. Otherwise, if one system goes down, you lose data in that cluster. 

Another issue is performance. You have to make sure you have the right configurations; otherwise, it will lead to queuing where all your jobs get queued.

It is a scalable product. 

You only get technical support when you take the cloud subscription. If you have it in-house, you won't get any support. If you have a cloud subscription, then the support is pretty good. You can raise a ticket from the UI, and they will respond within 24 hours.

So, the support team is pretty good but there is a little room for improvement. 

Neutral

The initial setup is pretty difficult since we deployed it in-house. We didn't use the cloud subscription, so we have to handle the deployment very carefully.

The challenge was deploying it and having the replication concept working. Another challenging feature is persistent volumes. You have to make sure the data is available on all clusters; otherwise, if one cluster goes down, you'll lose all your data. It's better to have it replicated.

We first used the cloud subscription, but we saw a possibility to reduce costs, so we tried deploying the open-source ClickHouse on-premises. That saved us money, but we didn't get all the features that come with the subscription.

We did it in-house.

Pricing for the cloud version is alright, not very costly or cheap. 

But if you have an in-house deployment on Kubernetes or something, it's going to be very cheap since you'll be managing everything.

I would tell other users to do a POC because it depends upon the business use case and the data. They can explore first. There's another open-source option called Apache Druid, which is a little better than ClickHouse. If that doesn't fit the use case, then they could go for ClickHouse.

Overall, I would rate the solution a seven out of ten. 

If you have a real-time basis, you should take a look at ClickHouse because it works on a vector database, and the querying is super fast compared to traditional databases. So, if your use case is real-time or logging or real-time dashboarding, then ClickHouse is a tool to consider. Otherwise, if it's batch processing and you can expect some latency, then you should go for other databases.

Hybrid Cloud

During my experience Clickhouse was used primary for companies analytics. At the same time I had a chance to apply it for various other use-case, such as log and storage metrics, Clickhouse as an ETL engine, monitorin, alerting and many more

ClickHouse is a user-friendly solution that tries to be compatible with SQL standards. It also tries to provide command-line tools, very nice formatting, libraries. And of course blazing speed it is main selling point of this technology

I would like ClickHouse to work more on integration with third-party tools. The solution has a lot of integrations, but most of them are not completed or production-ready.

The solution's setup requires some work and understanding. People at the company do not fully understand how to use ClickHouse. They try to use it like any relational database, which causes a lot of problems.

I have been using ClickHouse since 2018

We have experienced small bugs many times with the solution. Five years ago, the tool wasn't so stable. Now, it's better, but bugs still happen.

I rate the solution a seven out of ten for stability.

ClickHouse is a very scalable solution because it is designed to be scalable out of the box. There are some issues with scalability because, by design, ClickHouse does not support data rebalancing. I work at a start-up where around ten users use the solution.

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

I have never interacted with the solution's technical support because I usually use the open-source version of ClickHouse. You can post your issue on GitHub at any time, and you will usually get a response.

Positive

Currently, ClickHouse provides a cloud-based solution that you can use in the cloud. ClickHouse has an open-source version, which is free to use and has almost all the features.

ClickHouse provides the best real-time query performance in the market if it is used properly. Generally, ClickHouse is not so easy to use because it's designed in such a way that you should be aware of the infrastructure. The solution is complicated, but it is easy for someone with experience and knowledge. I would recommend the solution to other users.

ClickHouse is a magnificent solution, but users should first read a few articles about it to understand how to use it properly and how not to use it. Users should take a learning course to be aware of its architecture and to use it properly.

Overall, I rate the solution ten out of ten.