You are on page 1of 9

Pinterest using Amazon Web Services

A cloud computing platform

Multidisciplinary course - Report


Nikita Nair - 23060223174
INTRODUCTION
OVERVIEW ON PINTEREST AND PURPOSE

Pinterest is a visual discovery engine that hosts billions of images for over 450
million users to explore, save, and share as “Pins” to personalized digital
inspiration boards. Born in the Amazon Web Services (AWS) Cloud, Pinterest can
scale processing, storage, and analysis of its rapidly increasing data, all while
reducing infrastructure management and focusing on innovation.

Using compute solutions from AWS, Pinterest migrated its iOS continuous
integration and continuous delivery (CI/CD) pipeline from on-premises to reduce
build failures by over 80 percent. Pinterest also uses machine learning (ML) to
power its visual search tool Pinterest Lens that can recognize over 2.5 billion
objects and match them to products. Today, Pinterest’s exabyte data platform runs
entirely on AWS, scaling log search and analytics to over 1.7TB while reducing
operations costs by 30%.

Amazon Web Services (AWS) offers a highly secured on-demand cloud computing
platform that includes database storage, content delivery and other services to help
organizations scale and grow. Trusted by millions of customers, AWS allows them
to leverage its cloud products and solutions to build complex applications that are
highly scalable, flexible and reliable in nature.
WHY DID I CHOOSE PINTEREST?

1.Relevance

Pinterest is a widely recognized platform with millions of users worldwide.


Exploring how Pinterest utilizes AWS can be relevant and interesting to a broad
audience, especially those interested in technology, cloud computing, and social
media platforms.

2.Technical Insights

Presenting on Pinterest's architecture on AWS provides an opportunity to delve


into technical details such as their use of AWS services like EC2, S3, DynamoDB,
and others, as well as their strategies for load balancing, data storage, and content
delivery.
3.Trend Analysis

By examining Pinterest's journey on AWS, one can also identify trends in cloud
adoption, scalability challenges, and the evolving landscape of technology
infrastructure.Overall, choosing Pinterest on AWS as a presentation topic offers a
blend of practical insights, technical depth, and real-world relevance.

THESIS

This thesis explores Pinterest's strategic utilization of Amazon Web Services


(AWS) across various facets of its infrastructure, including storage, data solutions,
machine learning, migration, cost optimization, and security. Through a
comprehensive analysis, the study delves into Pinterest's innovative approaches to
leveraging AWS services such as Amazon Simple Storage Service (S3), AWS
Identity and Access Management (IAM), Amazon S3 Glacier Deep Archive,
Amazon Elastic Compute Cloud (EC2), and AWS Secure Token Service (STS).

The research highlights how Pinterest has successfully accelerated development


times, ensured always-on access for users, powered machine learning engines for
visual search, improved pipeline reliability, optimized costs, and enhanced security
measures through scalable access control solutions. By examining Pinterest's
collaboration with AWS and the resulting enhancements in its platform, this thesis
sheds light on best practices and lessons learned for organizations seeking to
maximize the benefits of cloud computing services for their own technological
innovations and operational efficiencies.

USE OF CLOUD COMPUTING


HOW DOES PINTEREST USE AWS?

Pinterest utilizes AWS across various facets of its infrastructure to enhance


scalability, performance, and innovation. Firstly, AWS serves as the backbone for
Pinterest's storage solutions, with services like Amazon Simple Storage Service
(S3) and Amazon S3 Glacier Deep Archive used to store billions of images and
manage data for its visual discovery engine. Additionally, AWS powers Pinterest's
data solutions, ensuring always-on access for users through machine learning
engines.

The platform leverages AWS storage and compute solutions to conduct hundreds
of millions of visual searches each month, providing personalized
recommendations to users. Moreover, Pinterest has migrated critical systems like
its iOS build pipeline to AWS, resulting in significant reliability improvements.

The company's commitment to cost optimization is evident as it scales log search


and analytics while reducing costs on Amazon OpenSearch Service. Security is
also paramount, with Pinterest implementing scalable access control solutions and
token vending services using AWS Secure Token Service (STS) and AWS Identity
and Access Management (IAM). Overall, Pinterest's strategic utilization of AWS
empowers it to scale its infrastructure, enhance operational efficiency, and deliver
an optimal user experience to its vast user base.

WHAT CLOUD IS IT?

Amazon Web Services (AWS) prides itself on being architected as the most
flexible and secure cloud computing environment available today. The core
infrastructure of AWS is meticulously designed to meet the stringent security
requirements of various entities, including the military, global banks, and other
high-sensitivity organizations. This commitment to security is underscored by a
comprehensive array of cloud security tools, comprising over 300 services and
features dedicated to security, compliance, and governance. Moreover, AWS boasts
support for 143 security standards and compliance certifications, ensuring that
customers can confidently operate within a framework that adheres to
industry-leading security protocols. Through its robust security measures and
commitment to innovation, AWS continues to set the standard for secure cloud
computing environments, empowering organizations to leverage the cloud with
peace of mind.

BENEFITS AND CHALLENGES


BENEFITS OF USING AWS ON PINTEREST

1. STORAGE

Pinterest Uses AWS to Create a System to Speed Up Development Times


using Amazon Simple Storage Service and AWS Identity and Access
Management (IAM) Pinterest Worked with AWS to Create a New Way to
Manage Data Access with Amazon S3 Pinterest Uses Amazon S3 Glacier
Deep Archive to Manage Storage for its Visual Discovery Engine

2. DATA SOLUTIONS

Pinterest Engaged with AWS Proserve to Ensure Pinners have Always-On


Access to Its Platform.

3. MACHINE LEARNING

Pinterest builds on AWS storage and compute solutions to power the ML


engines behind the Lens camera feature on its app, which is used to conduct
hundreds of millions of visual searches each month.

4. MIGRATION

Pinterest Improves iOS Build Pipeline Reliability by 80.5% Using Amazon


Elastic Compute Cloud Mac Instances.

5. COST OPTIMIZATION

Pinterest Scales Daily Log Search and Analytics from 500 GB to 1.7 TB,
Reduces Costs by 30% on Amazon OpenSearch Service.

6. SECURITY

Scalable access control solution for clusters of Amazon EC2 instances that
need access to data in Amazon S3. Pinterest built a token vending service
using AWS Secure Token Service (STS) that helps to bridge its internal
concept of users and authentication to managed policies and AWS Identity
and Access Management (IAM).

7. SUSTAINABILITY

Sustainability is integrated into Pinterest’s AWS architecture decisions and


how AWS continues to innovate on chip design as the organization works
toward Amazon’s goal of achieving net-zero carbon by 2040.
CHALLENGES

1. In their endeavor to enhance the responsiveness and dynamism of Pinterest


products, the team encountered several challenges related to real-time data
storage, high-performance machine learning ranking, and efficient candidate
generation for content recommendations. Despite having data stored in
MySQL and HBase clusters, the latency in querying these data stores for
online applications was impractical due to multiple round trips per request
and extensive data scanning. Additionally, there was a need for a
high-performance machine learning ranking system capable of processing
thousands of content pieces per request within a P99 latency of mere dozens
of milliseconds.

2. To address these challenges, the team embarked on building stateful services


capable of providing real-time updates and handling complex data models
efficiently. These services were essential for providing feature data to the
machine learning ranking system. Furthermore, a candidate generation
system was required to deliver high-quality candidate sets in real-time for
content recommendations. Throughout this process, the team made design
decisions, solved technical problems, and optimized systems to meet
stringent latency requirements for online applications.

3. In transitioning from Java to C++11, FBThrift, Folly, and RocksDB, the


team aimed to achieve low long-tail latency and handle significant fanout,
while catering to CPU-intensive tasks such as Scorpion, Pixie, and RealPin.
Despite initial challenges in setting up dependencies and build
environments, the adoption of C++ proved beneficial, facilitating the
development of more efficient systems across the organization.

4. The earlier architecture, predominantly relying on pregenerated content


stored statically in HBase, served Pinterest well in accommodating 100
million monthly active users by 2015. However, it posed limitations in
creating dynamic and responsive products, hindering experimentation and
real-time data utilization. Moreover, the storage cost associated with
pre-generating content for every user, including those who never returned,
was substantial.

To overcome these limitations and meet the demands of a dynamic and responsive
Pinterest, the team developed nine distinct systems. These systems revolutionized
the backend architectures supporting major Pinterest products, such as Following
Feed, Interest Feed, and Picked For You recommendations. While tailored for
Pinterest's needs, these systems offer solutions to common challenges faced by
web-scale consumer-facing content distribution applications.

CONCLUSION
SUMMARY

Pinterest relies on Amazon Web Services (AWS) to manage its vast data volume and
meet the demands of its rapidly growing user base. By leveraging AWS's scalable
storage solutions such as Amazon Simple Storage Service (S3) and Amazon S3
Glacier Deep Archive, Pinterest efficiently stores billions of images and manages
storage for its visual discovery engine. This architecture enables Pinterest to handle
its massive data volume while ensuring efficient access and management of content.
Additionally, AWS supports Pinterest's robust data solutions, allowing it to provide
always-on access for users through machine learning engines powered by AWS
storage and compute solutions. Features like Pinterest Lens, which conducts hundreds
of millions of visual searches each month, rely on AWS infrastructure to deliver
personalized recommendations to users. Moreover, Pinterest has successfully
migrated critical systems like its iOS build pipeline to AWS, resulting in significant
reliability improvements. This migration showcases how AWS services like Amazon
Elastic Compute Cloud (EC2) optimize operational processes and enhance system
performance. Furthermore, Pinterest's commitment to cost optimization is evident as
it scales daily log search and analytics while reducing costs by 30% on Amazon
OpenSearch Service. By leveraging AWS's flexible pricing models and cost-effective
solutions, Pinterest efficiently manages its infrastructure costs while maintaining
operational excellence. Security is also a top priority for Pinterest, as it implements
scalable access control solutions and token vending services using AWS Secure
Token Service (STS) and AWS Identity and Access Management (IAM). These
measures ensure secure access to data and authentication mechanisms, safeguarding
Pinterest's platform and user data against potential threats. Overall, Pinterest's
adoption of AWS empowers it to scale its infrastructure, enhance operational
efficiency, and deliver an optimal user experience to its vast user base, positioning it
as a leader in the dynamic landscape of social media platforms.

FUTURE
The future of cloud computing for systems like Pinterest on AWS holds immense
potential for further innovation and scalability. As technology advances, we
anticipate continued enhancements in cloud infrastructure to accommodate the
evolving needs of large-scale platforms like Pinterest. One key area of
development is in optimizing cloud services for real-time data processing and
analysis, enabling faster insights and personalized experiences for users.
Additionally, advancements in machine learning and AI capabilities within the
cloud will empower platforms like Pinterest to deliver even more intelligent and
tailored recommendations, driving user engagement and satisfaction. Furthermore,
we anticipate increased emphasis on sustainability and environmental
responsibility within cloud computing, with efforts to reduce energy consumption
and carbon footprint. Pinterest's commitment to sustainability, integrated into its
AWS architecture decisions, sets a precedent for future cloud-based systems to
prioritize eco-friendly practices. Overall, the future of cloud computing for
platforms like Pinterest lies in continuous innovation, enhanced scalability, and a
focus on sustainability, ultimately shaping a more dynamic and responsive digital
landscape for users worldwide.
BIBLIOGRAPHY
1. https://aws.amazon.com/solutions/case-studies/innovators/pinterest/
2. https://aws.amazon.com/what-is-cloud-computing/?nc1=f_cc
3. https://aws.amazon.com/what-is-aws/?nc1=f_cc
4. https://medium.com/pinterest-engineering/building-a-dynamic-and-responsi
ve-pinterest-7d410e99f0a9
5. https://www.linkedin.com/pulse/pinterest-case-study-aws-aditi-chaudhary
6. https://www.contino.io/insights/whos-using-aws
7. https://jinlow.medium.com/pinterest-system-architecture-fb49bd534424

● All keywords are highlighted.


● Headings are in bold underline and subheadings are in bold.
● Content is shared in the report and presentation.

THANK YOU

You might also like