With over 5 million users, and delivering over 16 billion emails per month, MailChimp is one of the most highly regarded Email Service Providers, especially among those focused on small and medium sized businesses. In order to serve customers effectively, MailChimp required a solution that could scale cost-effectively and execute email sends reliably for its expanding user base. When a colleague requested information regarding PowerMTA™, MailChimp responded with the following:

Reliability

We utilize PowerMTA form Port25, and this is what I can say… It’s reliable! The one recommendation I have to make sure you are running on bare metal and make sure you are using plenty of RAM. We license several instances of PowerMTA delivering well over 16B emails per month. We run 18 processors with 48GB of RAM in each box. With PMTA, make sure that the maximum amount of RAM it can use is LESS than the total RAM on the OS; otherwise, you will get into situations (when volume is really high) where the queues get wonky. Also, make sure that you limit your outbound bandwidth.

Bounce Categorization

The bounce categorization engine out of the box is improving. PowerMTA currently has over 17 different “baked in” categories, and users have the ability to define custom categories based on their own sending experiences. Boogietools is an excellent resource for bounce categorization as well. But if you have time and love regex’s, you can make PMTA’s bounce categorization work very well. Version 4.0 of PowerMTA will have even more powerful bounce categorization features.

Injection Rates

MailChimp has hours where we send about 2.5 million emails (sometimes more). We limit how many campaigns we have going simultaneously, but we inject 100,000 payloads an hour. We allow for up to 2.5 million recipients to be going out of an MTA at one time, and on high volume days we have 10 - 12 hours where we remain over 15 million every hour. PowerMTA can take what you throw at it but you don’t want to let the queues get overwhelmed. So, you need to use some logic and common sense and back off when you are reaching 65 - 70% of the MTA’s total volume.

VirtualMTA™ Utilization

Port25’s VirtualMTA stuff is awesome. Authentication, configuration, backoff algorithms, pause queues, force queues to clear, etc. On the dozens of MTAs we’ve installed, we have a bunch of IPs on each box and it pretty much manages itself. The setup is very manual but in Version 4.0, their new web interface will simplify most of this stuff. And the configuration stuff is straightforward.

Honestly, I cannot say enough good stuff about their software; it’s unbelievable for the price. But what really keeps us going with these guys is their support. They always, always have an answer and they really know their stuff.