By Josh Koenig October 13, 2014
Over the past two months we have upgraded our entire fleet — the actual hardware that powers websites on the Pantheon Platform — to faster CPUs and SSD disks. Faster websites win because better user experiences lead to more repeat visits and conversions.
These improvements help every Pantheon customer succeed. As Artur Bergman, VP of Engineering at Wikia and Fastly said, "If you're not using SSDs, you're wasting your life."
Pantheon has always been synonymous with speed, and based on our monitoring, the upgraded fleet is delivering improved performance at 40% on average. Your specific mileage may vary, but if you’ve noticed things seem snappier lately, this is why.
We have been able to do a complete fleet upgrade without disruptions or any scheduled maintenance because of our next-generation architecture, and this is just one of many improvements we’ve made to enhance performance and stability. Stay tuned for more updates on the performance front this fall.
Why SSDs Matter
The impact from faster CPUs are obvious: doing calculations more quickly leads to faster responses. For SSDs, the benefits come from much-improved read/write performance on the hard drive, especially random reads. Because SSDs do away with a spinning disk and a mechanical read head that needs to seek out the data—basically the same way a record player needle needs to be placed in the right part of an album to play a certain bit of song—they pull information from different parts of the drive much more quickly.
This boosts performance on several levels. At the application tier, it improves the speed of code deployment and helps PHP more quickly cache interpereted application code. It also helps if there is a need to examine many files on disk as part of an admin operation.
There’s also a serious performance boost for the database. While it’s always possible that you can code your way into a “query of death” for your application, SSDs will help speed up marginally inefficient queries quite a bit—shaving valuable milliseconds off page render times—and can help recover more quickly from a truly brutal query, should that ever happen.
Better Resiliency = Uptime Improvements
There are also significant benefits for Pantheon internally. We recently had every single host get a reboot from our infrastructure provider because of a serious security patch to the Xen hypervisor. It’s like the whole platform did a barrel roll:
While this was still a challenge—uncontrolled restarts across hundreds of hosts is going to make a dent in uptime no matter what— the fact that we made the switch to SSDs in advance meant we were back online much more quickly. We clocked an average of less than 10 minutes to recover from a full host system reboot, as opposed to 20 minutes or more with spinning disk systems.
This is what you get from a professional website platform. This fleet upgrade is just one of many examples of how Pantheon keeps you on the cutting edge of performance. Stay tuned for more updates on the performance front.
Topics: Speed & Performance
