PayPal partners with Google Cloud

[Music] Good afternoon. At PayPal, we’ve been able to move 15% of our infrastructure fleet into the Google Cloud, and we are just getting started. Today I want to walk you through the journey that weíve been on, but before I do that, I want to tell you a bit about PayPal. Most of you know us, but we are a group of companies: PayPal core, Venmo, Braintree, Zoom, and most recently, iZettle. Together, what we are attempting to do is to make sure that we can connect the world with payments. We are a two-sided network. On the one hand, we have more than 200 million consumers, and on the other hand, we have about 18 million merchants, and for a total of 237 million active users. Last year, we did about four hundred and fifty-six billion dollars in payments, and our mission on the one hand is to power digital and mobile commerce and on the other hand to democratize financial services around the world. Let me walk you quickly through the technology stack we have at the company. At the bottom-most is our infrastructure layerówhat we do internally in terms of building our data centers, our network, our hardware engineering teams, our storage databases and ops teams. On top of that, we have a layer of technology services that we use internally within the company to help subsidize software development across PayPal, but at the same time create some level of homogeneity and standardization as we build applications and services. Up from there is the payments operating system. This includes identity, payments, risk, compliance, and credit. And finally, the heart of PayPal, which is the experiences, and these experiences are the ones for merchants on the one hand and for consumers on the other side. The most important thing we do at PayPal is, of course, security. This is the number-one thing. We are in the business of trust, and security is the number-one thing that we need to build for. To give you an idea of the scale, as I mentioned, we did 456 billion dollars in payments last year. We operate in about 200 different markets, about a hundred currencies, and last year the peak day was about twenty- nine million payments on a single day. This translates to about fourteen thousand four hundred and fifty dollars per second on average. Just imagine the peaks. To power all of this, we have 2,700 different application services which are built and maintained by about 4,500 engineers across the world. Last year, in 2017, we did 17,000 software releases. Our internal infrastructure is about two hundred thousand servers, about 27 megawatts, and we have 238 petabytes of data all growing at about 30 percent year-over-year. The cloud opportunity for us to begin with was all about the developer and test environments. In 2015, we made the decision that we want to move to the public cloud and start to leverage what is now a 10-year-old industry. In doing so, we wanted to first learn from pilots, and we picked the developer and test environment as our first pilot. The reasons that we moved to the public cloud for are as follows: Number one is scale. As I mentioned to you, the number of transactions we take across the number of countries we have and the number of customers that we serve is very large, and the number of payments at PayPal is growing at about 25 percent every single year. Second, we wanted to make sure that we were accessible and present for our customers in their locale, in their geography and region. For this, we couldn’t go and build data centers all over the place. We needed a partner that could do this with us. Second was flexibility. We have, we are a payments business, and we are available everywhere around the world. And with mobile commerce, there is huge variations in the amount of payments that you get at any point in time of the day or any day of the week or any week of the year. To work with this, we needed a flexible infrastructure that could scale up and down to our demands and our needs. Third was the fact that we had to meet regulatory obligations. We are in 200 different markets. Each one of these markets is a jurisdiction. These jurisdictions have laws and money, and we need to be compliant with all of those. In some cases, we need to process and store data locally across various countries in the world. And finally, in terms of efficiencyóone, to understand what kind of a fleet we have; number two, to continuously optimize how much we are using; and number three, to ensure that we burst off when we need something and we don’t have hardware on premises that is not going to get used for a long time. To give you an example, our peak days for payments are in December. Once we capacitize for those payments, we don’t need that capacity till later in the first quarter of the year. With a public cloud, we have the ability to flex up and down. And finally, the rate of innovationópublic cloud gave us the ability for us to get started on the modernization journey and to ensure that we’re using the latest and greatest feature set that is available to make sure that our services are of the highest reliability, the highest security, and are providing the latencies that our customers need around the world. The reasonóthis is what I started with thoughóit was important that as we went into the transformation, we had to convince all the people internally within the company on the reason why we are doing this. This is the main reason why we moved to the public cloud. So our journey, like I mentioned, started in 2015 with discovery, and through 2016 we picked four cloud environments and ran a bunch of pilots to actually compare them side-by-side. At the end of that test, we picked the Google Cloud to host our applications in. We then started our journey in the beginning of 2017. January of 2017 is when we started the project to take all of the dev and test environment and move them to the Google Cloud. Through that journey in six months, we had an amazing partner with Google, and we learned iteratively through every sprint we went through. And on June 28th of 2017, we cut over dev and test over a weekend to the Google Cloud. This was about 15% of our infrastructure fleet. Now when we did this cut over, it was not all rainbows and unicorns. We had some challenges, and I want to share that with you as well. And those challenges: Number one, with respect to security. As a company that is focused on trust, our biggest need was to make sure that we were as secure as we were on-prem. In doing so, one, we wanted to ensure through tests that we had the risk mitigation and the controls required in the Google Cloud Platform and with the partnership with Google we got comfortable with this and started to move in that direction. Second and most important thing was culture. In any transformation, there are three pillars: there is a change in mindset, a change in the tool set, and the change in skill set. Most people start with the tool set. In this case, the Google Cloud was a tool set, but if you don’t change how you work every day and how you think of your application services in its structure, those tools will not help you in the long term. Third was architecture. Lift and shift was not going to work. We had to change our applications to make sure that they were ready for the public cloud, so we went through a process of ensuring that as we were making this migration, we were using the best of what the public cloud had to offer and not rely only on the paradigms that we had built on in the past. We also had a few challenges of the features. Even though public cloud is a 10-year- old industry, Google was still building many of the features. But through the partnership, we worked on various bridge technologies to help with our journey and that wait was well worth it. And since then, Google has built those features for us and we’ve incorporated that into our system. So taking you back to the journey, I mentioned how in the mid of 2017 we moved the dev and test environment into the public cloud and with that 15%, and since then we’ve had a ton of new learnings. We have now run this environment for a whole year and in doing so we have now qualified many other use cases that we can start to move to production. As we built the … beginning to build out to the production zones, we now have an intake process for the use cases. We are allowing for the development teams across the company to bring in the use cases that they want to migrate, qualifying that those use cases, both from a security perspective, a technical reliability perspective, and also from an efficiency perspective. And we intend in the next year to take a reasonable amount of our web and mid-tier and move that to the Google Cloud as well. In doing all this, we’ve had amazing learnings, and Google Cloud has been a tremendous partner. We have challenged each other, and I think we have learned from each other in this process. But we feel like the journey is just getting started, and we will continue to make progress towards putting more workloads in the Google Cloud. Thank you. [Applause] [Music]

Leave a Reply

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