Amazon Prime Day 2019 - Powered by AWS
The TestMace team continues to publish translations about interesting aspects of the web sphere. Next in line is a description of Amazon Prime Day in terms of developing highly loaded services. Enjoy reading!
By tradition, I would like to tell you about some of the advantages of AWS that helped us launch such a large-scale event and cope with a high load. From the How AWS Powered Amazon's Biggest Day Ever and Prime Day 2017 - Powered by AWS articles, you can find out how we processed the results of past Prime Day sales and what we did to optimize our systems and processes.
This time I would like to dwell on three features of AWS that helped us cope with record levels of traffic and sales during Prime Day: video processing infrastructure, database infrastructure and computing infrastructure. Let's get to know each of them better.
On July 10, Amazon Prime users were able to watch the Prime Day Concert , hosted by Amazon Music. This event was broadcast live, and the list of stars included artists such as Dua Lipa , SZA , Becky G and headliner in the person of the owner of 10 Grammy statuettes Taylor Swift .
Conducting a live broadcast of this scale and complexity to an audience of more than 200 countries requires careful planning and powerful infrastructure. Our Amazon Prime Video colleagues used various AWS media services to encode and package the video stream, including AWS Elemental MediaPackage , as well as AWS Elemental Live encoders.
For broadcasting, two regions were involved with a backup pair of data processing pipelines in each of them. The pipelines transmitted 1080p 30fps video to various content delivery networks (including Amazon CloudFront ) and did it as efficiently as possible.
To ensure high availability and stable operation of the system during an extremely high load during the Prime Day sale, both relational databases and NoSQL were used.
Amazon DynamoDB supports several high-traffic sites and systems, including Alexa, Amazon.com, and all of Amazon's 442 fulfillment centers. Within 48 hours of sales, these resources made 7.11 trillion requests to the DynamoDB API , setting a maximum value of 45.4 million requests per second.
Amazon Aurora is also used to run Amazon fulfillment centers. During Prime Day, 1900 database instances processed 128 billion transactions, saved 609 terabytes and transferred 306 terabytes of data.
Prime Day 2019 also included a large collection of diverse EC2 instances. The internal metric for these instances is called server equivalent, and during the sale, the numbers ranged from 372 thousand equivalents to a maximum of 426 thousand. EC2 instances used large volumes of the Elastic Block Store (EBS). On the eve of Prime Day, 63 petabytes of memory were added to the storage, and as a result, 2.1 trillion requests were processed per day and 185 petabytes of data were transferred.
I have given you impressive numbers that show what level of scalability you can achieve using AWS. As you can see, with AWS you can easily add resources to your system for conducting one-time (or periodic) events, even if these events will be broadcast to the whole world, and then return everything to its original state.
If you are thinking about holding a global event, I advise you to pay attention to those posts that I indicated above, as well as master the AWS Infrastructure Event Management program. My colleagues are always ready to help you with the launch of a large-scale software product, infrastructure migration or marketing event. This is how they work with them:
By tradition, I would like to tell you about some of the advantages of AWS that helped us launch such a large-scale event and cope with a high load. From the How AWS Powered Amazon's Biggest Day Ever and Prime Day 2017 - Powered by AWS articles, you can find out how we processed the results of past Prime Day sales and what we did to optimize our systems and processes.
This time I would like to dwell on three features of AWS that helped us cope with record levels of traffic and sales during Prime Day: video processing infrastructure, database infrastructure and computing infrastructure. Let's get to know each of them better.
Amazon Prime Video Infrastructure
On July 10, Amazon Prime users were able to watch the Prime Day Concert , hosted by Amazon Music. This event was broadcast live, and the list of stars included artists such as Dua Lipa , SZA , Becky G and headliner in the person of the owner of 10 Grammy statuettes Taylor Swift .
Conducting a live broadcast of this scale and complexity to an audience of more than 200 countries requires careful planning and powerful infrastructure. Our Amazon Prime Video colleagues used various AWS media services to encode and package the video stream, including AWS Elemental MediaPackage , as well as AWS Elemental Live encoders.
For broadcasting, two regions were involved with a backup pair of data processing pipelines in each of them. The pipelines transmitted 1080p 30fps video to various content delivery networks (including Amazon CloudFront ) and did it as efficiently as possible.
AWS Database Infrastructure
To ensure high availability and stable operation of the system during an extremely high load during the Prime Day sale, both relational databases and NoSQL were used.
Amazon DynamoDB supports several high-traffic sites and systems, including Alexa, Amazon.com, and all of Amazon's 442 fulfillment centers. Within 48 hours of sales, these resources made 7.11 trillion requests to the DynamoDB API , setting a maximum value of 45.4 million requests per second.
Amazon Aurora is also used to run Amazon fulfillment centers. During Prime Day, 1900 database instances processed 128 billion transactions, saved 609 terabytes and transferred 306 terabytes of data.
Amazon Computing Infrastructure
Prime Day 2019 also included a large collection of diverse EC2 instances. The internal metric for these instances is called server equivalent, and during the sale, the numbers ranged from 372 thousand equivalents to a maximum of 426 thousand. EC2 instances used large volumes of the Elastic Block Store (EBS). On the eve of Prime Day, 63 petabytes of memory were added to the storage, and as a result, 2.1 trillion requests were processed per day and 185 petabytes of data were transferred.
To summarize
I have given you impressive numbers that show what level of scalability you can achieve using AWS. As you can see, with AWS you can easily add resources to your system for conducting one-time (or periodic) events, even if these events will be broadcast to the whole world, and then return everything to its original state.
If you are thinking about holding a global event, I advise you to pay attention to those posts that I indicated above, as well as master the AWS Infrastructure Event Management program. My colleagues are always ready to help you with the launch of a large-scale software product, infrastructure migration or marketing event. This is how they work with them:
All Articles