Atlassian’s two-year cloud journey

A few years in the past, Dropbox shocked lots of people when it determined to principally drop the general public cloud, and built its own datacenters. Extra not too long ago, Atlassian did the alternative, closing most of its datacenters and shifting to the cloud. Corporations make these selections for quite a lot of causes. When Atlassian CTO Sri Viswanath got here on board in 2016, he made the choice to maneuver the corporate’s greatest functions to AWS.

Partly, it is a story of technical debt — that’s the idea that over time your functions turn out to be encumbered by layers of crusty code, making it more durable to replace and ever more durable to keep up. For Atlassian, which was based in 2002, that invoice got here due in 2016 when Viswanath got here to work for the corporate.

Atlassian already knew they wanted to replace the code to maneuver into the long run. One of many causes they introduced Viswanath on board was to steer that cost, however the pondering was already in place even earlier than he acquired there. A small staff was shaped again in 2015 to work out the imaginative and prescient and the structure for the brand new cloud-based method, however they needed to have their first CTO in place to hold it by way of to fruition.

Shifting to microservices

He put the plan into movement, giving it the interior code identify Vertigo — perhaps as a result of the considered shifting most of their software program stack to the general public cloud made the engineering staff dizzy to even think about. The objective of the mission was to rearchitect the software program, beginning with their greatest merchandise Jira and Confluence, in a such a method that it will lay the inspiration for the corporate for the subsequent decade — no strain or something.

Photograph: WILLIAM WEST/AFP/Getty Photos

They spent an excellent a part of 2016 rewriting the software program and getting it arrange on AWS. They focused on turning their 15-year outdated code into microservices, which in the long run resulted in a smaller code base. He stated the technical debt points had been very actual, however they needed to be cautious to not reinvent the wheel, simply change what wanted to be modified each time doable.

“The code base was fairly massive and we needed to go in and do two issues. We needed to construct it for multi-tenant structure and we needed to create microservices,” he stated. “If there was a service that might be pulled out and made self-contained we did that, however we additionally created new companies as a part of the method.”

Migrating clients on the fly

Final yr was the migration yr, and it was certainly a full year-long mission emigrate each final buyer over to the brand new system. It began in January and resulted in December and concerned shifting tens of hundreds of shoppers.

Photograph: KTSDesign/Science Photograph Library

To start with, they automated no matter they may they usually additionally had been very deliberate when it comes to the migration order, being aware of migrations that could be tougher. “We had been considerate in what order emigrate. We didn’t wish to do best first and hardest on the finish. We didn’t wish to do exactly the more durable ones and never make progress. We needed to mix [our approaches] to repair bugs and points all through the mission,” he stated.

Viswanath said that the overarching objective was to maneuver the purchasers with out a main incident. “For those who speak to anybody who does migration, that’s an enormous factor. Everybody has scars doing migrations. We had been aware to do that fairly fastidiously.” Surprisingly, though it wasn’t excellent, they did handle to finish the whole train with out a main outage, a degree of which the staff is justifiably proud. That doesn’t imply that it was at all times clean or simple.

“It sounds tremendous simple: ‘we had been considerate and we migrated,’ however there was warfare every single day. While you migrate, you hit a wall and react. It was a day by day factor for us all year long,” he defined. It took a complete staff effort involving engineering, product and help. That included having a buyer help individual concerned within the day by day scrum conferences so they may get a really feel for any points clients had been having and repair them as shortly as doable.

What they gained

As in any cloud mission, there are some normal advantages to shifting an software to the cloud round flexibility, agility and useful resource elasticity, however there was greater than that when it got here to this particular mission.

Photograph: Ade Akinrujomu/Getty Photos

To start with it has allowed sooner deployment with a number of deployments on the similar time, due largely to the copious use of microservices. Which means they’ll add new options a lot sooner. In the course of the migration yr, they held off on new options for probably the most half as a result of they needed to maintain issues as static as doable for the shift over, however with the brand new system in place they’ll transfer rather more shortly so as to add new options.

They get significantly better efficiency and in the event that they hit a efficiency bottleneck, they’ll simply add extra sources as a result of it’s the cloud. What’s extra, they had been in a position to have an area presence within the EU and that improves efficiency by having the functions nearer to the top customers situated there.

Lastly, they really discovered the cloud to be a extra economical choice, one thing that not each firm that strikes to the cloud finds. By closing the datacenters and lowering the capital prices related to shopping for {hardware} and hiring IT personnel to keep up it, they had been in a position to cut back prices.

Managing the individuals elements

It was a protracted drawn out mission, and as such, they actually wanted to consider the human side of it too. They’d swap individuals out and in to verify the engineers stayed recent and didn’t burn out serving to with the transition.

One factor that helped was the corporate tradition on the whole, which Viswanath candidly describes as one with open communication and a normal “no bullshit” coverage. “We maintained open communication, even when issues weren’t going nicely. Individuals would elevate their hand in the event that they couldn’t sustain and we might get them assist,” he stated.

He admitted that there was some anxiousness inside the firm and for him personally implementing a mission of this scale, however they knew they wanted to do it for the way forward for the group. “There was undoubtedly nervousness on what if this mission doesn’t go nicely. It appeared the plain proper course and we needed to do it. The danger was what if we screwed up in execution and we didn’t understand advantages we got down to do.”

In the long run, it was numerous work, but it surely labored out simply advantageous they usually have the system in place for the long run. “Now we’re arrange for the subsequent 10 years,” he stated.


Facebook Comments
Subscribe to our newsletter
Sign up here to get the latest news, updates and special offers delivered directly to your inbox.
You can unsubscribe at any time