dreamtech11

[dreamtech11]

Scaling for Success: How Dream11 Uses Predictive Analytics and Real-time Monitoring to Ensure 100% Uptime During Peak Seasons(Part-3)Alternate Version

Becoming scale management ready for the future

In accordance with our data science strategy, we perform predictive scaling. The model forecasts the expected concurrency for each match. This forecast and our load test served as the foundation for our benchmarking of our figures to scale over 100 microservices. In the past, we manually employed ASG scales and evaluated them on a regular basis, which was a difficult and time-consuming operation. The procedure is now automated, and our monitoring technology makes data verification quicker. To assist us detect and address problems, we have set up a number of notifications.

We feed predicted capacity statistics to our monitoring system from both a cloud watch and our benchmark for an hour. We can identify provisioning problems more quickly by comparing the data.

The scaling method adjusts to avoid a bottleneck whenever there are unpredictably high spikes during a postponed match toss, unexpected weather, or when a crucial player is not announced for the lineups. As a result, the network can be scaled up or down automatically to accommodate changing traffic demands. Often, all of this occurs just before the game begins.

As a team, we enjoy trying new things and frequently add new features. We developed Team Sharing & Guru Teams for the T20 World Cup, allowing players to use both expertly built teams as well as teams they could share with friends. Also, we tried both the KnockOut and Gladiator game forms, and the users gave us very positive comments on both. We streamlined the calculation for the group leaderboard and made it quicker to update after each match because people enjoy competing against their friends and family during tournaments. We added the eagerly anticipated emoji reactions to Chat Messages as an additional surprise.

What's coming

The tech team is currently working to prepare the systems for the Indian Premier League (IPL) in 2023 by incorporating the lessons learned from the most recent competition. IPL will be more popular in the next year than it was in 2022, therefore Dream11 systems must be more durable, dependable, and strong to guarantee the best possible user experience.

Related Tags

New App technologies

How Dream11 works

Latest technologies

IT industry challenges

Monitoring tool

Read full blog at

https://tech.dream11.in/blog/power-of-machine-learning-ahead-of-big-matches

[dreamtech11]

Scaling for Success: How Dream11 Uses Predictive Analytics and Real-time Monitoring to Ensure 100% Uptime During Peak Seasons(Part-2)

Using technology to address some of the most difficult IT industry challenges

Efficiency is enhanced by being observable from a single dashboard.

Since our network is complicated and scattered among microservices, monitoring is essential to ensuring accelerated diagnosis. We can monitor the performance of the DNS, application, infrastructure, and ingress/egress architecture with the use of our monitoring tools. Prior to the round lock, every second counts. If not handled properly, it can take time and effort, especially since many factors need to be taken into account, from networking to applications and business performance metrics. During a fantasy sports competition, our status pages and dashboards assist us in concentrating on the areas that need immediate attention.

Examples include the top Relational Database Service (RDS) depending on the connection formed, the top Application Programming Interface (API) with a response time of greater than 200 ms, or the Central Processing Unit (CPU). On a single dashboard, we have built a bird's-eye view of the whole Dream11 infrastructure. It allows us to resolve problems rapidly and reduce the Mean Time To Detect (MTTD) and Mean Time To Resolution (MTTR). Our monitoring tool can establish connections between logs, network measurements, Cloudwatch metrics, and APM metrics.

Performance Benchmarking & Testing

Another critical phase of any software's life cycle is performance testing. To identify flaws and create standards for each technical component, we conduct routine chaos and load testing.

All of our new app technologies in our apps contain clever user handling to make sure that even when backend applications perform poorly, the user experience is not adversely affected.

We can immediately detect network issues by using our network monitoring. Checking the TCP retransmits by AZ is a clear indication that this is true. To match the performance of our network, we provide a variety of slice-and-dice choices.

The availability zones, services, ENVs, domains, hosts, IP, VPCs, ports, regions, IP type, etc. may all be filtered, including traffic from local, private, and public IP addresses.

For instance, our APM product offers distributed end-to-end tracing from frontend hardware to databases. Our monitoring tools lets us automatically monitor service dependencies and latency to remove problems for our users to obtain the best experience possible by seamlessly correlating dispersed traces with frontend and backend data. We can solve the issue of delivering visibility to a request's lifecycle across several systems by using distributed tracing.

This is quite helpful for debugging and determining the areas of the programm where the greatest time is spent. We have a service map that examines each service to determine its RED metrics, dependencies, and filtering options for Application Services, Databases, Caches, Lambda Functions, and Custom Scripts. Nearly in real-time, the monitoring agent delivers data to our tool every 10 seconds, and this service map reflects this very instant. The map displays all services in green if no difficulties are found and in red if any are. This information is retrieved from the monitor set up for each service.

[dreamtech11]

Predicting traffic patterns for a platform like Dream11, with 150 million sports fans and over 10,000 contests, can be challenging, especially during peak seasons like IPL and World Cup, where traffic can spike from thousands to millions of concurrent users in just a few minutes. Let's learn how Dream11 tackles this issue using monitoring tools, performance testing and latest technologies.

[dreamtech11]

Scaling for Success: How Dream11 Uses Predictive Analytics and Real-time Monitoring to Ensure 100% Uptime During Peak Seasons(Part-1)

At Dream11, we strive to offer the best sports engagement experience for our users. As the largest fantasy sports platform with 150 million fans participating in over 10,000 contests, it can be difficult to predict traffic patterns, especially during high-demand events like the IPL and World Cup. To maintain 100% uptime during these matches, we use prediction-based scaling and scale-out infrastructure.

We generate an immense amount of data from our mobile applications and microservices, capturing user action events, system metrics, network metrics, and more. Our machine learning algorithms process this data to predict demand based on factors such as players, tournaments, virality, and user playing patterns. During the T20 ICC World Cup, for example, our platform is capable of managing up to 6.21 million concurrent users at the edge layer.

At Dream11, our Dreamsters play a crucial role in ensuring an optimal user experience. Our service owners have readiness lists and runbooks in place to quickly resolve any incidents, and our customer service team is equipped to handle incidents efficiently.

The observability of our network through a single dashboard helps us stay efficient, and monitoring through monitoring tools is critical for accelerated troubleshooting. Our monitoring tools track the performance of our infrastructure, application, and DNS, and our status pages and dashboards allow us to quickly identify and address issues. Our unified dashboard provides a bird's eye view of the entire Dream11 infrastructure, reducing Mean Time To Detect (MTTD) and Mean Time To Resolution (MTTR).

Overall, it's evident that Dream11 is using technology to tackle some of the biggest IT industry challenges, and it's great to see the emphasis on delivering an exceptional user experience.

[dreamtech11]

Scaling for Success: How Dream11 Uses Predictive Analytics and Real-time Monitoring and how they use the monitoring tool to Ensure 100% Uptime During Peak Season.