3. Today’s Agenda
Examine common needs and challenges
we’ve seen from Media and Entertainment
customers.
Provide examples of how customers have
met these needs using AWS.
Dig in on patterns for
Storage, Processing, and
Distribution.
Learn how to get started.
4. Media Storage
• Library storage for processing
• Archiving
Media Processing
• Encode/Transcode
• Encryption, watermarking
• Applying individualization and monetization mechanisms (advertising, PPV,
subscription, purchase, etc.)
Media Distribution
• Audio and video streaming of Live and VOD content
• Managed distribution to set top boxes
• Over the top (OTT) distribution
• Internet distribution
• Rich media web applications
5. Media Industry Trends and Needs
• Media industry is at an inflection point
• Infrastructure requirements growing exponentially
• As the media industry has “gone digital”, content distribution
requirements have also evolved
• Scalable and secure media storage, processing and distribution
• Anytime, anywhere, any device consumption
• Personalized content and experiences
• Expectation of low latency, global distribution
• Support multiple business models (purchase, PPV, subscription)
• Media has unusual performance, security and reliability
characteristics compared to other industries
6. Today’s Ubiquitous Media is Complex and Challenging
High Performance Computing
Server Farms
Content Distribution Networks
DAM, DRM, Encoding, E
diting, and Video Player
Software
Large Number of File Formats
Massive Storage Networks
7. On-Premise Infrastructure is Costly & Complex
Large Capital Expenditures Underutilized IT Assets
Software Maintenance
Out of Datacenter Space
Budget Constraints
Manual Deployments
Contract negotiation
Limited Capacity
Expensive IT products
“
Difficulties managing physical growth
IT spends 80% of its time and
resources keeping the lights on
8. The Cloud Provides a Better way
Instant Elasticity Rapid Time to Value
Focus on
Innovation not
On-Demand
Infrastructure
Pricing
No CapEx
9. How the Cloud Can Reduce Digital Lifecycle
Complexity
Massive Scale
Rapid File Transfer
Lower IT Costs
High Reliability
Accredited Security
10. Massive Scale Makes it Easier to Hit Deadlines
Problem: Timeline overruns to
support PS3 launch
Solution: AWS’s Scale, 1200 virtual
machines on-demand
Benefits:17,000
titles (80TB
of data) transcoded in
days, able to support
launch
11. Massive Scale Makes it Easier to Focus on Your
Core Business
Problem: Massive global growth in online
gaming
Solution: AWS’s Global Infrastructure
Benefits: Scaled to handle 4 million
players in 8 weeks
12. Storage
S3 – Storage for Source / Mezzanine files
Exceptional Durability: 99.999999999%
Secure
Economical
EBS – Storage for Media during Processing
POSIX Block Device
Dynamic
Persistent for non-contiguous processing
Direct Connect – Leverage your existing storage systems via a low latency
Nx10Gbps link to AWS
13. S3 Use Cases and Users
• Media Serving
• Media Sharing
• File Storage / Backup
• Static Content
• Big Data
15. Moving Files to and From Amazon S3 is Fast
Symantec NetBackup
S3n Internet
Cloudberry Pro
Riverbed Whitewater
Multi-part Upload Up to 700 Mb/s direct 1Gb and 10 Gb
to Amazon S3 Connections
per node
16. New! Storage Gateway
Your Datacenter
Amazon Elastic
Compute Cloud
(EC2)
AWS Storage
Gateway
Clients
VM SSL
Internet
On-premises Host or
Direct AWS Storage Amazon Simple
Connect Gateway Service Storage Service (S3)
Application
Servers Amazon Elastic
Block Storage
(EBS)
Direct Attached or Storage Area Network Disks
17. Moving Files Faster Also Makes it Easier
to Hit Deadlines
9 Minutes: 7 Seconds:
Other than Amazon S3
Amazon S3
18. Processing
EC2 – Servers in the Cloud by the Hour
High-Performance, Scalable, Secure
Origin servers, custom application servers, monetization and DRM systems
EMR – A coordination framework for distributed computing
Minimizes Administration
Automates distribution
Simplifies workflows
CG1 – CUDA – GPGPU-enabled instances for super high FP performance, transcoding.
20. Easily Scale Up and Down
Internet Video App on Amazon EC2
From 50 to 5,000 servers in 3 days
5,000
Scaled to peak of
Number of EC2 Instances
5,000 instances
in 3 days
Launch of
Facebook
application
0 Monday Tuesday Wednesday Thursday Friday Saturday Sunday
The Animoto Blog
21. AWS Reduces IT Costs
Saved 70%
OUTAGE
OPPORTUNITY
COST
Saved 50%
Traditional
Hardware
Actual
Demand
Automated
Virtualization
Reduced costs by over 50%
22. Why Vimeo Chose AWS
“
Our users don't care whether it's a
peak time or not, and we prefer not
“
to pay for transcoding machines
the five days a week we aren't
using them.
This workload is perfect for EC2
23. CG1 – GPGPU is here
= 10x faster, 2x higher res, lower cost?!
24. Distribution
Cloudfront – Next-Gen Content Delivery Network
Performance-oriented
Multi-format
Secure
Economical
Route 53 – DNS
API controlled, allows programmatic shifts
Weighted Records
Alias records for ELB
EC2– Origin servers, custom application servers,
monetization and DRM systems
25. CloudFront in a nutshell
Our Content Delivery Network Delivers:
• Low latency. Improves media load times.
• High bandwidth. Enables high bit rate HD video and
other media applications.
• Redundant. Eliminates single points of failure.
• Scalable. Ensure great experience as number of end
users grows.
• Global. Worldwide network provides great experience
regardless of geography.
• Cost-effective. Pay as you go model provides flexibility
for your business.
25
26. Delivery of Non-Cached Video Download
Request
S3 bucket or Object routed to most
custom origin with requested CloudFront optimal edge
your stored video
content
from origin Edge
location 0 End-user requests
2 Locations
1 video
Media stored
3 in cache
4
Data transfer of 5 End-user enjoys
Data transfer of video to video
CloudFront edge cached video
location to cache download to end user
27. Delivery of Cached Video Download
S3 bucket or Request routed
custom origin with CloudFront to most optimal
your stored video
content
Edge
Locations
edge location 0 End-user requests
video
1
Skipping these
steps means
lower latency for
your users!
Media
read from
cache 2 of
Data transfer 3 End-user enjoys
video
cached video
download to end user
29. New Feature!
CloudFront
for Dynamic
Websites
• Deliver your entire site from
CloudFront
• Fast Performance using
reliable AWS network
connections
• Simple, Self-Configuration
using API or our Management
Console
• Cost-Effective with no
minimum commits or required
long-term contracts
30. • Different origins – same
distribution
• Persistent TCP Connections
• Large (10) TCP Window
• Full support for Query Strings
• 0 second minimum TTL
31. AWS’ Reliable Infrastructure Makes it
Easier to Focus on Other Tasks
“As with all the AWS services we leverage, using Amazon
CloudFront is so simple and reliable that the team doesn’t
have to think about it. It all just works, freeing us to focus on
building cool applications.”
>1Pb/mo through Cloudfront
32. AWS and CloudFront’s global presence
AWS Regions
US East (Northern Virginia)
US West (Northern California)
US West (Northern Oregon)
SA East (Sao Paulo)
Europe (Dublin)
Asia Pacific (Singapore)
Asia Pacific (Toyko)
AWS CloudFront Locations
United States: New York, NY (2) Europe: Asia: South America:
Ashburn, VA Palo Alto, CA Amsterdam Hong Kong Sao Paulo
Dallas, TX San Jose, CA Dublin Tokyo
Jacksonville, FL Seattle, WA Frankfurt Singapore
Los Angeles, CA (2) South Bend, IN London Osaka!
Miami, FL St. Louis, MO Paris
Newark, NJ Stockholm
Milan!
32
33. What do M&E customers expect from a CDN:
• Great performance to a global audience.
• Reliable delivery to wide range of clients.
• Scalability for unpredictable spikes in demand
• Control for secure content delivery
• Support for HD high-bandwidth streaming
• Reporting and analytics to track & analyze viewing patterns
• Easy to configure and manage solutions
• Cost-effective solutions
33
34. Great performance to a global audience.
• Amazon is a metrics
driven company.
Gomez Large Object Test
• We focus on metrics Japan
that capture the end 12.000
Download Time (Seconds)
9.673
10.000
user experience: the 8.000
“last mile,” not internet 6.000
3.688 3.441
“backbone” data
4.000 3.126
2.000
centers. 0.000
Last Mile
• Your customers do CloudFront Comparison A
not live in data Comparison B Comparison C
centers. Last mile data based on 7,907 observations taken between 17-SEP-2011 and 1-Oct-2011
*The Gomez tests were designed and conducted by Amazon using the Compuware Corporation performance
network. The test results have not been reviewed, approved or endorsed by Compuware Corporation
35. Reliable delivery to wide range of clients.
• Multiple delivery protocols for Adobe FMS 4.5
different platforms and devices
• Adobe RTMP
• HTTP Streaming for iOS
?
• Microsoft Silverlight
• Options for live and on-demand video
• Full control over origin for live
streaming
• Reliability backed by CloudFront
Service Level Agreement
35
36. Example: Video banner ad customer.
80
70
Delivered (Gbps)
Volume of Data
60
Peak usage over 60 Gbps
50
40
30
20
10
0
8:00 AM 9:00 AM 10:00 11:00 12:00 PM 1:00 PM 2:00 PM 3:00 PM 4:00 PM 5:00 PM 6:00 PM 7:00 PM 8:00 PM 9:00 PM
AM AM
36
37. secure content delivery controls
• Private Content Feature authenticate users with signed URLS
• Uses policy-driven access controls for control and flexibility
• Restrict on resource or path, time, source IP
• Signatures generated using asymmetric encryption
• SSL delivery and RTMPE streaming encrypt bytes on the wire
• Origin Access Identities secure your content in Amazon S3
• Identify and Access Management (IAM) to control who can
configure your CloudFront distributions
38. • Self service signup and configuration –
anytime, from anywhere
• AWS Management Console to create
and manage AWS infrastructure like
S3, EC2, CloudFront
• REST APIs for integration into your own
systems and workflow
38
39. Getting Started
• Simply sign up for AWS at
http://aws.amazon.com/
• Store some content in S3, put a CloudFront distribution on
it, and compare!
• Try your media processing applications on EC2
• Take advantage of the Free Tier to experiment with more
advanced services
Notes de l'éditeur
In today’s highly connected world where consumers typically have multiple devices that they consume video content on, consumer’s access to content in an on-demand “where you want it, when you want it, how you want it” world is really quite frictionless, at least from their point of view. But supporting frictionless consumption from a technology and scale perspective is more difficult than ever.We’ve listed some of the technical challenges here around the need for HPC, server farms, DAM, encoding, encryption, editing, and others, but the biggest issue is that you need to be able to do this at scale due to the large number of files that get processed.
The answer is the cloud!Cloud computing removes many of the physical limitations related to media processing. The scale of cloud resources can change the problem from a linear one to non-linear and enable companies to optimize how they process and distribute media in ways that have never before been available. Cloud computing:Gives you the agility to respond to opportunities more quickly than competitors.Enables efficient use of resources and allows you to focus on your core competencies and adding real value vs. managing infrastructureIn these times of tight budgets, it turns the conversation from capex to opex. Hardware depreciates quickly. For unpredictable demand, a rental model maximizes efficient use of resources.Moving to a service oriented model with utility billing enables you to deal with unexpected workloads via elastic scaling, with a pay-as-you-go model. And it optimizes the efficiency by handling the right tasks in the right place at the right timeExamples:Encoding can be done on cluster GPU instances to maximize speed or Spot instances to maximize budget. Enables you to make businesses decision vs. being beholden to technical or capacity limitations.In the future we expect additional focus on efficient use of resources for media processing and delivery. For example, from an efficiency perspective, once content is in the cloud, all media processing can happen while minimizing file movement.Get ProRes/mezzanine files into cloud storage, then have various processes act on those files.Also, there is a network effect: As content providers put more content into the cloud, more partners will move to a model where they can participate in the value chain.
Over the next few slides I’m going to talk about some of the ways that Cloud Computing can help, including…
Proof Point of how scales solves the growth problem (Netflix)Netflix wanted to make it’s library of content available to Sony Playstation 3 users when the SP3 launchedDue to timeline overruns, Netflix’s window of time to transcode their library of 17,000 videos was rapidly closing before the PS3 launch date.Netflix decided to make use of AWS’s massive scale and started 1200 EC2 instances to do the encoding.In just a few days Netflix was able to transcode over 80TB of video in time for the PS3 launch. This is a great example of how cloud computing such as Amazon Web Services can enable non-linear solutions to problems that traditionally could only be solve in a linear fashion
Because their games are often based on social networking, the demand can grow significantly in a short period of time, sometimes quite unexpectedlyThe beta launch of Restaurant City saw the audience grow to more than 4 million players in the first 8 weeks. They initially projected between 100k to 250k players…Able to scale their infrastructure seamlessly while keeping everyone’s concentration on their core business.
Before, taking advantage of AWS’s massive scale was limited by transfer speeds in to Amazon S3 and file size limitations.At Amazon we really focus on listening to our customers and innovating on their behalf, so we launch large object support , up-to 5TB, and Multi-part uploadAspera has created technology that takes advantage of multi-part uploads to transfer data to S3 at up-to 700 Mb/s (you will likely get a chance to see this live later today).We also launched Amazon Direct Connect, so companies can connect directly to AWS with 1Gb and 10Gb ports, bypassing the internet.
The ability to move files faster is possible due to our low-latency internal transfer. This is an efficiency gained by having content and processing both in the AWS cloud infrastructure. As an example, one of our encoding partners, Zencoder, has an encoding service that runs on top of AWS.In order to process the files, they must first be transferred to Zencoder. Those files can live on your own hard drive, a file server, or cloud based storage.Example is a 1.5 GB file to Zencoder for processingAs you can see, transferring files that already live in Amazon Web Services’ S3 storage is much faster than transferring files from storage locations outside of AWS…nearly 80X faster!
This is a graph of IT capacity over timeIn a traditional IT situation where you run the HW on premise, you can see how you would typically increase capacity in very discrete steps, likely aligning with when you get your new annual budget allocationsThe problem is that actual demand doesn’t usually follow that pattern. As you can see, there may be points in time where you’ve got excess capacity going unused – which is an opportunity cost, and at other times you may not have enough capacity – which leads to a poor customer experienceWith automated scalability of Amazon Web Services, your capacity can match your demand as it grows and shrinksWith On-Demand, Reserved, and Spot EC2 instances, you can choose what works best for you.Low cost: reduced pricing 15 times in last 5 yearsWe’ve removed the S3 data ingress cost as of July 2011Tiered pricing, rates go down as volume increasesFlexible: Pay-as-you-go pricing with or without commitments Transparent pricing: you get what you pay for
AWS’ Reliable Infrastructure Makes it Easier to Focus on Other Tasks. An example customer that has seen the benefits here is PBS.PBS is using Amazon’s CloudFront Content Delivery Network for its streamed media.They were using another CDN initially, but were having reliability issues so they wrote a monitoring tool to test other CDNs, including CloudFront.What they found was that CloudFront had a significantly lower error rate than the incumbent CDN, so within a matter of weeks they made the switch.PBS is now delivering over 1PB/mo through CloudFront.