Lessons on Building Analytic Applications with Altinity.Cloud
2021 has been a great year for Altinity.Cloud. Cloud users have grown to almost a third of our total subscription base. During that time both we and our customers have absorbed lessons on using ClickHouse effectively in the cloud. So what have we learned? In this article, I would like to call out three points for anyone working on real-time analytics and ClickHouse.
The only performance benchmark that counts is your own
Database vendors generate constant sound and fury over benchmark results. The latest shattered lances come from Snowflake and Databricks. To be honest, we’re sometimes guilty of excessive performance pride ourselves. We love ClickHouse and don’t hesitate to publish numbers that show it in a good light.
But when it comes to customer applications, only one thing matters: how fast your app runs on your data.
As soon as you have even the outlines of an application, it’s a great idea to test it in Altinity.Cloud. As my colleague John Hummel explains, spinning up a ClickHouse data warehouse is just a matter of a few clicks. And Alexander Zaitsev writes, there are abundant tricks to load data in a hurry. Finally, there’s no need to speculate about the effect of different cluster configurations. Just try them, as Alexander Zaitsev did with AWS Graviton testing.
Over the course of the last year, many dozens of customers followed this simple recipe. We’re happy to report that they succeeded, as Locale.ai wrote in their blog about how they scaled analytic performance to billions of rows per second.
A cloud console does not an analytic application make
I’m still in awe of Redshift, a pioneering service that combined cloud ease-of-use with a performant data warehouse. Prior to Redshift, it could take months even to try out a data warehouse. Redshift reduced the “time to try” to under 20 minutes. Yet for all its greatness, you still have to write the application.
The same is certainly true for ClickHouse. I have heard more than one user describe ClickHouse as their secret weapon. That’s true, but only in experienced hands. Development teams do not necessarily have that experience. They also need to focus on making applications work quickly rather than ClickHouse details.
This is why we couple outstanding ClickHouse support with Altinity.Cloud. The difference is that with a cloud platform, we don’t have to spend time debugging operational issues. Instead, we can focus on assisting developers to get apps out, as we did with GraphCDN. We helped them get ClickHouse analytics into production in 2 months from start to finish. As we documented in our article about that experience, the key was collaboration. One of our Slack threads ran to 89 messages over a few hours.
Another thing is that we are constantly testing and improving Altinity.Cloud integration with key analytic components like Kafka, Grafana, Superset, and Tableau. In the days of yore, everyone had to find and fix such software by themselves. Altinity.Cloud gives us at Altinity more time for eliminating the risk and time to integrate the parts into applications.
Get to market fast without compromising the future
It’s one thing to get an analytic application to market in a startup. Speed is everything. It’s different when your baby grows up and runs on dozens or even hundreds of servers. Cost margins, control of data, and long-term IT strategy become a lot more important.
It’s difficult to predict the future, so the simplest approach is to pick technology that allows you the widest range of choices for future deployments. In the case of analytic applications, that’s permissively licensed open source software that can run anywhere.
This is why Altinity constantly stresses that customers can move from Altinity.Cloud to self-managed ClickHouse (and vice versa). We are committed to the following concrete goals regarding open source.
- Uphold ClickHouse Apache 2.0 licensing.
- Use Apache 2.0 licenses on our own infrastructure like the ClickHouse Operator and ClickHouse Backup.
- Focus on integration with other open source tools like Kafka, Redpanda, Superset, Grafana, Rudderstack, and Debezium.
We also use Kubernetes as our operating platform for Altinity.Cloud and release container builds for all base infrastructure. We are committed to ensuring users can run them on any Kubernetes platform.
As we showed in recent blog articles, you can stand up ClickHouse clusters easily in Altinity.Cloud. You can also build equivalent clusters in your own Kubernetes environments. And it’s not just ClickHouse clusters. If you use open source infrastructure, you can move the entire analytic pipeline there as well. “Freedom to change” is key to analytic applications that can evolve as your business grows and the world changes around you.
It’s been our pleasure during 2021 to enable dozens of new customers to build innovative analytic applications on Altinity.Cloud. I’ve worked on databases and their applications for almost four decades. The wonder of seeing creative new ways to use data never goes away. We’ll continue to enable cutting-edge analytics and share what we learn with all ClickHouse users.
If you have enjoyed our recent posts on Altinity.Cloud, you can join the movement yourself. Sign up for a free trial today!
Originally published on the Altinity Blog on December 15, 2021.