There are a handful of providers offering Kafka as a Service. If you are in the market for managed Kafka you might be wondering what factors to consider when choosing a provider. In this post, we break down the five most important considerations.
#1 Is the service fully managed?
If the service is truly a fully managed solution double check that the company’s terms include the following:
- Ensuring high availability
- Real-time monitoring and alerting
- Provisioning servers
- Configuring Kafka manually
- Replacing servers if/when they fail
- Deploying server patches as needed
- Built in safeguards for supporting load changes and data spikes
- Performing upgrades
#2 What’s the uptime guarantee?
Do not engage with a managed Kafka provider if they don’t offer an uptime guarantee. The minimum uptime that I recommend to companies is 99.95%. For a 30-day month, that comes out to a possible 22 minutes of downtime. At least one managed Kafka as a Service provider, Dattell, offers a 99.999% uptime guarantee. That reduces the possible monthly downtime to less than a minute. Guarantees aside, our team is always working to keep our clients running at 100% with no downtime.
#3 Will my Kafka cluster(s) be secure?
Choose a provider that will meet your security needs/goals. This should include encryption, authorization, and authentication.
Additionally, avoid opening your sensitive data up to the internet by keeping your managed Kafka solution on your private network. Ensure only you control access to your data by having your authentication infrastructure, rather than a third party, control access to your data.
#4 What is the latency?
Many Kafka as a Service providers require that your Kafka clusters are built in their clouds. However, it’s typically best to have your Kafka cluster built alongside your infrastructure in your network to reduce latency. If your Kafka cluster is built alongside your infrastructure, that will result in tens of milliseconds of latency rather than hundreds-to-thousands.
#5 How does the Kafka as a Service provider profit?
Kafka as a Service providers that are hosting your Kafka cluster(s) in their clouds are incentivized to create larger than necessary, inefficient clusters to increase their profits. Additionally, they typically markup the cloud instance fees. Choose a provider that will fully manage Kafka on your servers or cloud instances for a more cost effective solution.