Editor’s observe: Flywheel, a WP Engine firm that focuses on managed WordPress internet hosting, wanted extra strong infrastructure for the platform it makes use of to run websites for over 35,000 companies and inventive manufacturers. Transferring to Google Kubernetes Engine gave them the reliability and useful resource utilization they wanted; rearchitecting their utility as a multi-tier, cloud-native utility working on Google Cloud managed providers gave them new ranges of operational effectivity and helped to manage their prices. Learn on to listen to how they did it.

Flywheel makes a speciality of offering highly effective web site options to artistic companies and has constructed a platform that’s custom-configured and optimized for the WordPress CMS. By providing devoted assets, strong workflow instruments, staffing WordPress specialists, and specializing in the pace of our websites, we’re capable of present nice website efficiency and person expertise for our clients.

To realize nice website efficiency and nice person expertise, we now have to innovate in and round WordPress, because it evolves with customers’ expectations. In an more and more aggressive panorama the place digital is the brand new norm, our customers’ expectations of our platform have by no means been larger.

In our early days, we constructed a single tier, VM-based platform and {custom} orchestration instruments to function it (like many SaaS firms). Initially, this strategy labored, however as we grew, we skilled excessive operational overhead, excessive upkeep wants, low useful resource utilization, and an absence of agility resulting from robust couplings. 

We discovered ourselves spending extra time on feeding the factor we had constructed somewhat than supporting our clients with new options and merchandise. We additionally lacked the power to investigate the information inside our platform at scale. We knew we would have liked to evolve our platform to at least one with decrease operational overhead and the agility to shift rapidly in response to buyer wants.

As part of reengineering our structure, we determined to experiment with Google Kubernetes Engine (GKE), Filestore, Cloud SQL, and BigQuery on Google Cloud. We rapidly found some necessary benefits that allowed us to enhance efficiency, higher handle our prices, and achieve vital sensible analytics capabilities: 

  • Useful resource environment friendly Kubernetes apps: We have been very impressed by GKE’s resource utilization. We have been capable of run the nodes at >90% CPU load, thereby lowering our unused assets by 85%. As well as, GKE saved nodes working reliably at this degree with none points. On our earlier supplier, the VMs couldn’t take a excessive CPU load and would usually grow to be unreachable.

  • Excessive-performance, totally managed file storage: Filestore is a totally managed, NoOps service that’s built-in with GKE. As such, we have been capable of simply develop or shrink the filesystem based mostly on the wants of our clients, with out having to function it ourselves.

  • Totally managed relational database service for MySQL: With Cloud SQL, we have been capable of scale our databases effortlessly, with out compromising efficiency or availability. We additionally noticed a big discount in upkeep prices. 

  • Serverless, extremely scalable, and cost-effective information warehouse: We have been thrilled by the facility of BigQuery to investigate our information and the power to share the findings with a couple of clicks. By integrating BigQuery into our new structure, we now have the sensible analytics capabilities we want, with zero operational overhead.

We needed to make the most of GKE’s improved useful resource utilization and reliability, plus implement a number of managed providers to keep away from the toil of working all the tiers ourselves. To do this, nevertheless, we would have liked to implement some easy modifications to our structure.

Constructing a container-based, multi-tier utility

To begin, we redesigned our structure to make use of light-weight containers. We knew this may require vital guide refactoring and we needed to get it proper the primary time. We relied closely on the crew at Google Cloud to assist us design and validate this structure earlier than we began writing any code. Google’s experience close to creating and working cloud native software program was a key differentiator in selecting them as a accomplice. It was via this deep partnership and tight collaboration that we have been capable of construct and execute the correct technique and obtain our desired outcomes.

This design choice stemmed from the better useful resource utilization of container-based purposes. Our VM-based platform usually left a substantial quantity of CPU idle, as much as 85% in some circumstances. With container-based workloads working on GKE, we went from utilizing solely 30% of the CPU on common to 90%, a 200% enhance! 

From there, we adopted one of the necessary ideas in software program engineering: maintain it easy. Our aim was that each single WordPress website on our platform would function with most efficiency and reliability. To realize this, we primarily relied on three Google Cloud providers:



Leave a Reply

Your email address will not be published. Required fields are marked *