LSports provides real-time sports events data spanning 30 different sports across the world. These real-time data feeds include metrics such as scores, penalty cards and even the size of the audience. The data serves LSports customers, including magazines and online gaming sites, that leverage LSports capabilities to create real-time user experience.
Over the past year, the company’s real-time data consumption has grown with new and existing customers. Initially, LSports’ systems were hosted on a colocation data center, however this traditional rigid physical infrastructure was not flexible enough to support the company’s new and growing demand. Adding servers or memory involved lengthy ordering processes from the colocation vendor. Aside from limiting business growth, their inflexible infrastructure posed a challenge to their operations team when it came to quickly implementing backup and disaster recovery mechanisms to enhance service robustness.
LSports’ operations team leveraged Emind cloud experts’ deep knowledge and expansive capabilities to design a new cloud architecture for deployment and operation. The initial discussion they had was around choosing the best cloud infrastructure (IaaS) to serve demand and simultaneously maintain their infamous real-time performance capabilities. While taking part in Emind’s Cloud Architecture Workshop, the LSports operations team, was advised to move their production environment to Google Cloud Platform (GCP).
Challenges and Solutions
The Main Challenge – Performance & Scalability
In addition to updating sports’ events data in real-time, LSports’ servers receive dozens of API concurrent customer requests to download hundreds of megabytes of sports events data every second. As service traction increases, so do system performance degradations. This additionally results in cross customer performance interferences as new customers start to use the multi-tenant system.
It was clear that the underlying infrastructure needed to be replaced and the solution is a scalable cloud infrastructure.
Solution
Due to the need for a high level of performance, especially when it comes to fast provisioning and network connections, GCP was the best candidate. First Emind ran a proof-of-concept to understand more about the network connection’s performance between GCP and the colocation center. The outcome of the network speed tests showed a 10ms delay between the sites, which gave the green light for migration.
Second, Emind enabled autoscaling for the API serving instances. The autoscaling thresholds included the instances’ CPU and API calls per second. GCP autoscale enabled LSports to scale from a few servers to dozens according to demand. In addition, the colocation SATA disks were easily replaced by SSD disks, which dramatically enhanced system performance and scalability.
One of the major considerations with the move to the cloud was ensuring that network performance would be maintained. The 1G network cards in LSports’ colocation center servers proved to be limiting when it came to maintaining real-time performance. However, the move to GCP equipped LSports with various cloud instance size options with up to 4G network cards. This enables fast user data downloads and streamlined data from the colocation center to GCP. Today, the system collects data in real-time and streams it to GCP instances with minimal latency.
High Availability
As mentioned above, LSports’ SLA includes providing customers with data in real-time, which meant that data center operations were faced with the challenge of limited flexibility and a lack of fast resource provisioning. As a result, the operations team had to manually migrate applications and data between servers even in the case of server failures due to routine maintenance in the colocation center. With LSports’ rapid growth, they had to enhance service robustness along with existing backup and disaster recovery mechanisms.
Solution
By leveraging GCP’s global presence, LSports’ system was deployed across two zones, eliminating single zone failure risks. What’s more, GCP autoscaling enabled the distribution of compute nodes across the zones. LSports’ API servers were distributed accordingly so if a zone failed, the load could be automatically routed to the healthy zone and autoscaling would support the added demand in the specified zone. In addition, the system’s databases were stored both in GCP and in the colocation center. These sites are synced in real-time so data can be consumed at any point in time from either site.
The Benefits
“This 3-way partnership between our team, Emind and Google Cloud Platform is an important factor in the success of our cloud deployment. The migration process was seamless to users and the results are great. Today we have a scalable solution that can support our business growth while maintaining efficient IT operations in terms of efforts and costs.”
In just over three months, LSports’ cloud migration project, with the help of Emind experts, resulted in system scalability that ensured end user performance utilizing of Google Cloud Platform. In addition, LSports’ data platform transformed to become even more robust, efficient and scalable than ever before.
The past colocated environment lacked the scalability required to handle the large amount of API calls and data retrievals. Moving to a hybrid cloud configuration allowed LSports to enjoy both worlds including the public cloud’s elasticity as well as their colocation site for off-shore database backup and non-critical mission operations. In addition, the new deployment enabled LSports to distribute services on a global scale across multiple data centers. That is the initial phase of gradually making a complete move of data and applications to the cloud. With scalable deployment on GCP, LSports can sleep well while business grows.
