There is no perfect way to implement OpenStack in your company, but there are general principles that can guide you to a successful implementation.
One of the advantages of open source software such as OpenStack is the ability to download it, try it out and get a hands-on understanding of it without long interactions with vendor vendors or the need for lengthy internal pilot project approvals between your company and your company. -vendor.
But what happens when it's time to do more than just try out a project? How will you prepare the deployed system from source to production? How can you overcome organizational barriers to the adoption of new and transformative technologies? Where to begin? What will you do next?
Of course, much can be learned from the experience of those who have already deployed OpenStack. To better understand the patterns of adoption of OpenStack, I spoke with several teams that have successfully introduced this system to their companies.
MercadoLibre: Dictate of Necessity and Run Faster than a Doe
If the need is strong enough, then implementing an agile cloud infrastructure can be almost as simple as βbuild it and they will come.β In many ways, this is the experience that Alejandro Comisario, Maximiliano Venesio and Leandro Reox have had at MercadoLibre, the largest e-commerce company in Latin America and the eighth largest in the world.
In 2011, as the company's development department began the journey of decomposing their then-monolithic system into a platform of loosely coupled services connected via APIs, the infrastructure team saw a surge in the number of requests their small team needed to fulfill.
βThe shift happened very quickly,β says Alejandro Comisario, MercadoLibre Cloud Services CTO. βWe literally realized overnight that we couldnβt continue to work at such a rhythm without the help of some kind of system.
Alejandro Comisario, Maximiliano Venesio, and Leandro Reox, the entire MercadoLibre team at the time, began looking for technologies that would allow them to eliminate the manual steps involved in providing infrastructure to their developers.
The team set themselves more challenging goals, setting goals not only for immediate tasks, but also for the goals of the entire company: reducing the time to provide users with production-ready virtual machines from 2 hours to 10 seconds and eliminating human intervention from this process.
When they found OpenStack, it became clear that this is exactly what they were looking for. The rapidly evolving MercadoLibre culture allowed the team to move quickly into building the OpenStack environment, despite the relative immaturity of the project at the time.
βIt became clear that the OpenStack approach of researching, diving into the code, and testing the functionality and scaling coincides with the MercadoLibre approach,β says Leandro Reox. βWe were able to jump straight into the project, define a test suite for our OpenStack installation, and start testing.
Their initial testing on the second release of OpenStack identified several issues that prevented them from going into production, but the transition from the Bexar release to the Cactus release came at just the right time. Further testing of the Cactus release gave confidence that the cloud was ready for commercial use.
Launching into commercial operation and developers' understanding of the possibility of obtaining infrastructure as quickly as developers are able to consume it, determined the success of the implementation.
βThe whole company was hungry for a system like this and the functionality it provides,β says Maximiliano Venesio, Senior Infrastructure Engineer at MercadoLibre.
However, the team has been careful in managing developer expectations. They needed to make sure that the developers understood that existing applications would not be able to run on the new private cloud without changes.
βWe had to make sure our developers were ready to write stateless applications for the cloud,β said Alejandro Comisario. βIt was a huge cultural shift for them. In some cases, we have had to teach developers that keeping their data in the instance is not enough. Developers needed to adjust their thinking.
The team was attentive in training developers and recommended best practices for building cloud-ready applications. They sent emails, hosted informal learning lunches and formal training sessions, and ensured that the cloud environment was properly documented. The result of their efforts is this: MercadoLibre developers are now as comfortable developing applications for the cloud as they were developing traditional applications for the company's virtualized environments.
The automation they were able to achieve with the private cloud paid off, allowing MercadoLibre to dramatically scale up its infrastructure. What started as an infrastructure team of three supporting 250 developers, 100 servers, and 1000 VMs has grown into a 10-person team supporting over 500 developers, 2000 servers, and 12 VMs.
Workday: Creating a business case for OpenStack
For the team at Workday, a SaaS company, the decision to adopt OpenStack was more strategic than operational.
Workday's path to private cloud adoption began in 2013 when the company's management agreed to invest in a broad-based software-defined data center (SDDC) initiative. The hope for this initiative was to achieve more automation, innovate, and improve the efficiency of data centers.
Workday created its vision for a private cloud between the company's infrastructure, engineering, and operations teams, and an agreement was reached to launch a research initiative. Workday has hired Carmine Remi as Chief Cloud Officer to lead the change.
Rimi's first task at Workday was to expand the original business case to a large part of the company.
The cornerstone of the business case was increased flexibility when using SDDC. This increased flexibility would help the company realize its desire for continuous software deployment with zero downtime. The SDDC API was supposed to allow Workday's application and platform teams to innovate in ways that weren't possible before.
Equipment efficiency was also taken into account in the business case. Workday sets ambitious goals to increase the utilization rate of existing data center equipment and resources.
βWe found that we already have middleware technology that can take advantage of the private cloud. This middleware has already been used to deploy dev/test environments in public clouds. With a private cloud, we could extend this software to create a hybrid cloud solution. Using a hybrid cloud strategy, Workday can migrate workloads between public and private clouds to maximize hardware utilization, delivering business savings
Finally, Rimi's cloud strategy noted that simple stateless workloads and their horizontal scaling will allow Workday to start using its private cloud with less risk and achieve cloud maturity naturally.
βYou can start with a plan and learn how to manage a new cloud with a light workload, akin to traditional R&D that allows you to experiment in a safe environment,β suggested Rimi.
With a solid business case, Rimi evaluated several well-known private cloud platforms, including OpenStack, against a broad set of evaluation criteria that included each platform's openness, ease of use, flexibility, reliability, resilience, support and community, and potential. Based on their assessment, Rimi and his team chose OpenStack and set about building a commercial-ready private cloud.
Having successfully implemented its first viable OpenStack cloud, Workday continues to push for wider adoption of the new SDDC environment. To achieve this goal, Rimi uses a multi-pronged approach focused on:
- concentration on cloud-ready workloads, especially on stateless applications in the portfolio
- defining criteria and migration process
- setting development goals for migrating these applications
- communicate and educate groups of interested Workday users using OpenStack meetings, demos, videos and trainings
βOur cloud supports a variety of workloads, some in production, others in production readiness. Ultimately we want to migrate all workloads and I expect we will reach a tipping point where we will see a sudden flurry of activity. We are preparing the system piece by piece every day to be able to handle this level of activity when the time comes.
BestBuy: breaking taboos
Electronics retailer BestBuy, with annual revenues of $43 billion and 140 employees, is the largest of the companies listed in the article. And so, while the processes used by the bestbuy.com infrastructure team in provisioning an OpenStack-based private cloud are not unique, the flexibility with which they applied these processes is impressive.
To implement their first OpenStack cloud at BestBuy, Web Solutions Director Steve Eastham and Principal Architect Joel Crabb had to rely on creativity to overcome the many barriers that stood in their way.
The BestBuy OpenStack initiative grew in early 2011 from an effort to understand the various business processes involved in releasing the e-commerce site bestbuy.com. These efforts have revealed significant inefficiencies in quality assurance processes. The quality assurance process introduced significant overhead with each major site release, which took place two to four times a year. A significant portion of these costs were related to manually setting up the environment, reconciling discrepancies, and resolving resource availability issues.
To address these issues, bestbuy.com launched the On-Demand Quality Review initiative led by Steve Eastham and Joel Crabb to identify and address bottlenecks in bestbuy.com's quality assurance process. Among the main recommendations of this project were the automation of quality control processes and the provision of self-service tools to user groups.
And while Steve Eastham and Joel Crabb were able to use the prospect of very significant quality control costs to justify private cloud investment, they quickly ran into a problem: although the project was approved, the project was not funded. There was no budget for the purchase of equipment for the project.
Necessity is the mother of invention, and the team took a new approach to funding the cloud: they swapped the budget for two developers with another team that had a hardware budget.
With the budget received, they intended to buy the equipment needed for the project. Contacting HP, their hardware supplier at the time, they set about optimizing the offering. Through careful negotiation and an acceptable reduction in equipment requirements, they were able to cut equipment costs by almost half.
In a similar vein, Steve Eastham and Joel Crabb made a deal with the company's networking team, taking advantage of the existing core capacity while saving the typical costs associated with purchasing new networking equipment.
βWe were on pretty thin ice,β Steve Eastham said. βThis was not a common practice at BestBuy at the time, nor is it now. We were operating below the radar. We could have been reprimanded, but we managed to avoid it.
Overcoming financial difficulties was only the first of many obstacles. At that time, there was practically no opportunity to find OpenStack experts for the project. So they had to build a team from scratch by bringing together traditional Java developers and system administrators in the team.
βWe just put them in the same room and said, βFind out how to work with this system,β says Joel Crabb. βOne of the Java developers told us: βThis is crazy, you can't do it. I don't know what you're talking about."
We had to combine the different styles of the two types of commands to get the result we wanted - a software-driven, testable, incremental development process.
Incentivizing the team at an early stage of the project allowed them to get some impressive victories. They were able to quickly replace an outdated development environment, reduce the number of quality assurance (QA) environments, and in the process of transformation, they got the way new teams work and the speed of application delivery.
Their success has given them good opportunities to ask for more resources for their private cloud initiative. And this time they had support at the level of the top management of the company.
Steve Eastham and Joel Crabb secured the funding needed to hire additional staff and five new racks of equipment. The first cloud in this wave of projects was the OpenStack environment, which runs Hadoop clusters for analytics. And it is already in commercial operation.
Conclusion
In the MercadoLibre, Workday, and Best Buy stories, there are a number of principles that can guide you on your path to a successful OpenStack implementation: be open to the needs of developers, businesses, and other potential users; work within the established processes of your company; cooperation with other organizations; and be prepared to act outside the rules when necessary. All these are valuable soft-skills that are useful to have with the OpenStack cloud.
There is no perfect way to implement OpenStack in your company - the way to implement it depends on many factors related to both you and your company and the situation you are in.
While this fact may be confusing to OpenStack fans wondering how to implement their first project, it is nonetheless a positive point of view. This means there are no limits to how far you can go with OpenStack. What you can achieve is only limited by your creativity and resourcefulness.
Source: habr.com