I’m going to apologize in advance for the quality of my drawings in this blog post. My talents do not include being gifted in building beautiful diagrams and what you see here is a quick Visio knock-up. At its worst.
With this said, I’m writing this blog post today to tell you about some exciting new enhancements to our Private Email service that we’re working on.
In short, if you use and love our Private Email service, you’ll see dramatic improvements to performance and reliability. From May 2015 onwards, we’re going to be offering a 99.9% uptime guarantee with this service. This means if we fail to achieve this in any calendar month (excluding rare, scheduled maintenance) then you will be entitled to account credits.
If you’re yet to use our Private Email service then there’s never a better time to give it a try. Completely ad-free, completely private and secure, we’re investing heavily to make it the fastest, most reliable, most secure and most private platform around.
And if you’re interested in some of the tech ‘nitty gritty’, please read on.
With any large scale email platform, the biggest challenge is always the underlying storage platform. A service provider needs something performant, something scalable and something reliable. To date, we’ve used solutions from NetApp, from Tegile and from Dell’s Compellent range. And while each has had its pros and cons, none of these has ticked all of our boxes. Moving forward, we’re enhancing out platform by building out an object storage ring powered by Scality.
After much internal R&D, evaluation, comparison and design work, Scality came out top of the various storage solutions. Comprising of a fully redundant ring with 40 gigabits of connectivity between the ring and the hardware that sits above it, Scality delivers the performance – and the reliability – that we strive to deliver.
The diagram below offers some insight into how we’re designing this.
Above the Scality Ring, we have redundant Dell blades housed in M1000 chassis providing CPU cores and RAM for our VMWare virtualization layer. Dell M420 blades give us the latest Intel Xeon E5 version 3s and fast RAM. VMWare consumes this into our various application servers.
This offers a number of benefits over the current physical server setup, with redundancy at the hardware, virtualization and software layers.
This also allows for rapid scaling out of out platform as it grows.
Above VMWare, we have F5 BigIron load balancers. Previously, we used software load balancers but these hardware load balancers offer a much more intelligent load balancing approach. Sessions are smartly load balanced across all of our web server VMs and these also introduce the ability to better load balance email connections – from your desktop clients, mobile devices and others.
And finally, above our F5s will be our brand new, next generation converged network that powers a good amount of our Phoenix infrastructure. This offers fast, resilient connectivity to our global audience. More importantly, the F5s are capable of determining the fastest node in our environment to place your connection on; ensuring the highest possible quality of service and immediate fail-over in the event that a node goes unhealthy.
I’m sure the question you’re begging to ask is ‘when’. Well the good news is that we have all of the hardware in place. We’re working hard building this out, deploying our applications, testing, training and documenting. We’re finalizing our migration plan to upgrade our current users to our new platform. And we think, realistically, we’ll be fully on the new platform by the end of May.