Maintaining the Front Door to Netflix
Ben Schmaus, @schmaus
APIcon SF - 2014
Global Streaming Video
for TV Shows and Movies
More than 48 Million Subscribers
More than 40 Countries
Netflix Accounts for >34% of Peak
Downstream Traffic in North America
Netflix subscribers are watching more than 1 billion...
Netflix Accounts for >6% of Peak
Upstream Traffic in North America
Netflix subscribers are watching more than 1 billion ho...
Team Focus:
Build the Best Global Streaming Product
Three aspects of the Streaming Product:
• Non-Member
• Discovery
• Str...
Netflix API : Key Responsibilities
• Broker data between services and Devices
• Provide features and business logic
• Main...
Netflix API : Key Responsibilities
• Broker data between services and Devices
• Provide features and business logic
• Main...
APIs Do
Lots of Things!
Data Gathering
Data Formatting
Data Delivery
Security
Authorization
Authentication
System Scaling
Discoverability
Data Con...
Data Gathering
Data Formatting
Data Delivery
Security
Authorization
Authentication
System Scaling
Discoverability
Data Con...
Definitions
• Data Gathering
– Retrieving the requested data from one or many local
or remote data sources
• Data Formatti...
Meanwhile…
There are two players in APIs
API Provider
API Provider API Consumer
API Provider
PROVIDES
API Consumer
CONSUMES
Traditional API Interactions
API Provider
PROVIDES
EVERYTHING
API Consumer
CONSUMES
Everything means, API Provider does:
• Data Gathering
• Data Format...
Why do most API providers provide
everything?
• API design tends to be easier for teams closer
to the source
• Centralized...
Why do most API providers provide
everything?
• API design tends to be easier for teams closer
to the source
• Centralized...
Data Gathering Data Formatting Data Delivery
API Consumer
API Provider
Separation of Concerns
To be a better provider, the...
Data Gathering Data Formatting Data Delivery
API Consumer
Don’t care how data
is gathered, as long
as it is gathered
API P...
Data Gathering Data Formatting Data Delivery
API Consumer
Don’t care how data
is gathered, as long
as it is gathered
Each ...
Data Gathering Data Formatting Data Delivery
API Consumer
Don’t care how data
is gathered, as long
as it is gathered
Each ...
Lsud vs. sskd
Should you consider alternatives to
one-size-fits-all API model?
Ingredients:
• Small number of targeted API consumers is ...
• http://thenextweb.com/dd/2013/12/17/futur
e-api-design-orchestration-layer/
Because of our separation of
concerns, the Netflix API team is
enabled to focus on different charters
Brokering Data to
1,000+ Device Types
Screen Real Estate
Controller
Technical Capabilities
One-Size-Fits-All
API
Request
Request
Request
Courtesy of South Florida Classical Review
Resource-Based API
vs.
Experience-Based API
Resource-Based Requests
• /users/<id>/ratings/title
• /users/<id>/queues
• /users/<id>/queues/instant
• /users/<id>/recomm...
REST API
RECOMME
NDATIONS
MOVIE
DATA
SIMILAR
MOVIES
AUTH
MEMBER
DATA
A/B
TESTS
START-
UP
RATINGS
Network Border Network Bo...
RECOMME
NDATIONS
MOVIE
DATA
SIMILAR
MOVIES
AUTH
MEMBER
DATA
A/B
TESTS
START-
UP
RATINGS
OSFA API
Network Border Network Bo...
RECOMME
NDATIONS
MOVIE
DATA
SIMILAR
MOVIES
AUTH
MEMBER
DATA
A/B
TESTS
START-
UP
RATINGS
OSFA API
Network Border Network Bo...
Experience-Based Requests
• /ps3/homescreen
JAVA API
Network Border Network Border
RECOMME
NDATIONS
MOVIE
DATA
SIMILAR
MOVIES
AUTH
MEMBER
DATA
A/B
TESTS
START-
UP
RAT...
RECOMME
NDATIONSA
ZXSXX C
CCC
MOVIE
DATA
SIMILAR
MOVIES
AUTH
MEMBER
DATA
A/B
TESTS
START-
UP
RATINGS
JAVA API
SERVER CODE
...
RECOMME
NDATIONSA
ZXSXX C
CCC
MOVIE
DATA
SIMILAR
MOVIES
AUTH
MEMBER
DATA
A/B
TESTS
START-
UP
RATINGS
JAVA API
DATA GATHERI...
Because we are no longer
catering to LSUDs
We can now focus on building a
business on which all of Netflix
can operate
Netflix API : Key Responsibilities
• Broker data between services and Devices
• Provide features and business logic
• Main...
The bigger the ship…
the slower it turns
Distributed Architecture
1000+ Device Types
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
Reviews
A/B Test
Engine
Dozens of Dependenci...
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Dependency Relationships
2,000,000,000
Requests Per Day to the
Netflix API
30
Distinct Dependent
Services for the Netflix API
~600
Dependency jars Slurped
into the Netflix API
14,000,000,000
Netflix API Calls Per Day to
those Dependent Services
0
Dependent Services with
100% SLA
99.99% = 99.7%30
0.3% of 2B = 6M failures per day
2+ Hours of Downtime
Per Month
99.99% = 99.7%30
0.3% of 2B = 6M failures per day
2+ Hours of Downtime
Per Month
99.9% = 97%30
3% of 2B = 60M failures per day
20+ Hours of Downtime
Per Month
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Circuit Breaker Dashboard
Call Volume and Health / Last 10 Seconds
Call Volume / Last 2 Minutes
Successful Requests
Successful, But Slower Than Expected
Short-Circuited Requests, Delivering Fallbacks
Timeouts, Delivering Fallbacks
Thread Pool & Task Queue Full, Delivering Fallbacks
Exceptions, Delivering Fallbacks
Error Rate
# + # + # + # / (# + # + # + # + #) = Error Rate
Status of Fallback Circuit
Requests per Second, Over Last 10 Seconds
SLA Information
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Fallback
Personaliz
ation
Engine
User Info
Movie
Metadata
Movie
Ratings
Similar
Movies
API
Reviews
A/B Test
Engine
Fallback
Netflix API : Key Responsibilities
• Broker data between services and Devices
• Provide features and business logic
• Main...
Scaling the Distributed System
AWS Cloud
Autoscaling
Autoscaling
Scryer : Predictive Auto Scaling
Not yet…
Typical Traffic Patterns Over Five Days
Predicted RPS Compared to Actual RPS
More than 48 Million Subscribers
More than 40 Countries
Zuul
Gatekeeper for the Netflix Streaming Application
Zuul *
• Multi-Region
Resiliency
• Insights
• Stress Testing
• Canary Testing
• Dynamic Routing
• Load Shedding
• Security...
All of these approaches are
designed to prevent failures…
But sometimes the best way to
prevent failures is to force them!
I randomly
terminate instances
in production to
identify dormant
failures.
Chaos
Monkey
Netflix API : Key Responsibilities
• Broker data between services and Devices
• Provide features and business logic
• Main...
Dependency Relationships
Testing Philosophy:
Act Fast, React Fast
That Doesn’t Mean We Don’t Test
Automated Delivery Pipeline
Cloud-Based Deployment Techniques
Current Code
In Production
API Requests from
the Internet
Single Canary Instance
To Test New Code with Production Traffic
(around 1% or less of traffic)
Current Code
In Production
...
Canary Analysis Automation
Single Canary Instance
To Test New Code with Production Traffic
(around 1% or less of traffic)
Current Code
In Production
...
Current Code
In Production
API Requests from
the Internet
Current Code
In Production
API Requests from
the Internet
Current Code
In Production
API Requests from
the Internet
Perfect!
Stress Test with Zuul
Current Code
In Production
API Requests from
the Internet
New Code
Getting Prepared for Production
Current Code
In Production
API Requests from
the Internet
New Code
Getting Prepared for Production
Error!
Current Code
In Production
API Requests from
the Internet
New Code
Getting Prepared for Production
Current Code
In Production
API Requests from
the Internet
New Code
Getting Prepared for Production
Current Code
In Production
API Requests from
the Internet
Perfect!
Stress Test with Zuul
Current Code
In Production
API Requests from
the Internet
New Code
Getting Prepared for Production
Current Code
In Production
API Requests from
the Internet
New Code
Getting Prepared for Production
API Requests from
the Internet
New Code
Getting Prepared for Production
https://www.github.com/Netflix
Maintaining the Front Door to Netflix
Ben Schmaus, @schmaus
APIcon SF - 2014
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Maintaining the Front Door to Netflix
Upcoming SlideShare
Loading in …5
×

Maintaining the Front Door to Netflix

2,007 views

Published on

Talk about the Netflix API and how it serves as the front door for Netflix device UIs. Topics include: API design, resiliency patterns, scalability, and enabling fast dev/deploy cycles.

Published in: Internet, Technology, Education
0 Comments
4 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total views
2,007
On SlideShare
0
From Embeds
0
Number of Embeds
169
Actions
Shares
0
Downloads
0
Comments
0
Likes
4
Embeds 0
No embeds

No notes for slide
  • Netflix strives to be the global streaming video leader for TV shows and movies
  • We now have more than 44 million global subscribers in more than 40 countries
  • Those subscribers consume more than a billion hours of streaming video a month which accounts for about 33% of the peak Internet traffic in the US.
  • Those subscribers consume more than a billion hours of streaming video a month which accounts for about 33% of the peak Internet traffic in the US.
  • Our 44 million Netflix subscribers are watching shows and movies on virtually any device that has a streaming video screen. We are now on more than 1,000 different device types.
  • The subscribers can watch our original shows like Emmy-winning House of Cards.
  • Within this world, the Edge Engineering team focuses on these three aspects of the streaming product.
  • Most companies focus on a small handful of device implementations, most notably Android and iOS devices.
  • At Netflix, we have more than 1,000 different device types that we support. Across those devices, there is a high degree of variability. As a result, we have seen inefficiencies and problems emerge across our implementations. Those issues also translate into issues with the API interaction.
  • For example, screen size could significantly affect what the API should deliver to the UI. TVs with bigger screens that can potentially fit more titles and more metadata per title than a mobile phone. Do we need to send all of the extra bits for fields or items that are not needed, requiring the device itself to drop items on the floor? Or can we optimize the deliver of those bits on a per-device basis?
  • Different devices have different controlling functions as well. For devices with swipe technologies, such as the iPad, do we need to pre-load a lot of extra titles in case a user swipes the row quickly to see the last of 500 titles in their queue? Or for up-down-left-right controllers, would devices be more optimized by fetching a few items at a time when they are needed? Other devices support voice or hand gestures or pointer technologies. How might those impact the user experience and therefore the metadata needed to support them?
  • The technical specs on these devices differ greatly. Some have significant memory space while others do not, impacting how much data can be handled at a given time. Processing power and hard-drive space could also play a role in how the UI performs, in turn potentially influencing the optimal way for fetching content from the API. All of these differences could result in different potential optimizations across these devices.
  • Many UI teams needing metadata means many requests to the API team. In the one-size-fits-all API world, we essentially needed to funnel these requests and then prioritize them. That means that some teams would need to wait for API work to be done. It also meant that, because they all shared the same endpoints, we were often adding variations to the endpoints resulting in a more complex system as well as a lot of spaghetti code. Make teams wait due to prioritization was exacerbated by the fact that tasks took longer because the technical debt was increasing, causing time to build and test to increase. Moreover, many of the incoming requests were asking us to do more of the same kinds of customizations. This created a spiral that would be very difficult to break out of…
  • Many other companies have seen similar issues and have introduced orchestration layers that enable more flexible interaction models.
  • Odata, HYQL, ql.io, rest.li and others are examples of orchestration layers. They address the same problems that we have seen, but we have approached the solution in a very different way.
  • We evolved our discussion towards what ultimately became a discussion between resource-based APIs and experience-based APIs.
  • The original OSFA API was very resource oriented with granular requests for specific data, delivering specific documents in specific formats.
  • The interaction model looked basically like this, with (in this example) the PS3 making many calls across the network to the OSFA API. The API ultimately called back to dependent services to get the corresponding data needed to satisfy the requests.
  • In this mode, there is a very clear divide between the Client Code and the Server Code. That divide is the network border.
  • And the responsibilities have the same distribution as well. The Client Code handles the rendering of the interface (as well as asking the server for data). The Server Code is responsible of gathering, formatting and delivering the data to the UIs.
  • And ultimately, it works. The PS3 interface looks like this and was populated by this interaction model.
  • But we believe this is not the optimal way to handle it. In fact, assembling a UI through many resource-based API calls is akin to pointillism paintings. The picture looks great when fully assembled, but it is done by assembling many points put together in the right way.
  • We have decided to pursue an experience-based approach instead. Rather than making many API requests to assemble the PS3 home screen, the PS3 will potentially make a single request to a custom, optimized endpoint.
  • In an experience-based interaction, the PS3 can potentially make a single request across the network border to a scripting layer (currently Groovy), in this example to provide the data for the PS3 home screen. The call goes to a very specific, custom endpoint for the PS3 or for a shared UI. The Groovy script then interprets what is needed for the PS3 home screen and triggers a series of calls to the Java API running in the same JVM as the Groovy scripts. The Java API is essentially a series of methods that individually know how to gather the corresponding data from the dependent services. The Java API then returns the data to the Groovy script who then formats and delivers the very specific data back to the PS3.
  • We also introduced RxJava into this layer to improve our ability to handle concurrency and callbacks. RxJava is open source in our github repository.
  • In this model, the border between Client Code and Server Code is no longer the network border. It is now back on the server. The Groovy is essentially a client adapter written by the client teams.
  • And the distribution of work changes as well. The client teams continue to handle UI rendering, but now are also responsible for the formatting and delivery of content. The API team, in terms of the data side of things, is responsible for the data gathering and hand-off to the client adapters. Of course, the API team does many other things, including resiliency, scaling, dependency interactions, etc. This model is essentially a platform for API development.
  • If resource-based APIs assemble data like pointillism, experience-based APIs assemble data like a photograph. The experience-based approach captures and delivers it all at once.
  • And as we all know, the bigger the ship, the slower it turns. That was very much the case for Netflix years ago.
  • To grow to where we knew we needed to be, Netflix aggressively moved to a distributed architecture.
  • I like to think of this distributed architecture as being shaped like an hourglass…
  • In the top end of the hourglass, we have our device and UI teams who build out great user experiences on Netflix-branded devices. To put that into perspective, there are a few hundred more device types that we support than engineers at Netflix.
  • At the bottom end of the hourglass, there are several dozen dependency teams who focus on things like metadata, algorithms, authentication services, A/B test engines, etc.
  • The API is at the center of the hourglass, acting as a broker of data.
  • Our distributed architecture, with the number of systems involved, can get quite complicated. Each of these systems talks to a large number of other systems within our architecture.
  • Assuming each of the services have SLAs of four nines, that results in more than two hours of downtime per month.
  • And that is if all services maintain four nines!
  • If it degrades as far as to three nines, that is almost one day per month of downtime!
  • So, back to the hourglass…
  • In the old world, the system was vulnerable to such failures. For example, if one of our dependency services fails…
  • Such a failure could have resulted in an outage in the API.
  • And that outage likely would have cascaded to have some kind of substantive impact on the devices.
  • The challenge for the API team is to be resilient against dependency outages, to ultimately insulate Netflix customers from low level system problems and to keep them happy.
  • To solve this problem, we created Hystrix, as wrapping technology that provides fault tolerance in a distributed environment. Hystrix is also open source and available at our github repository.
  • To achieve this, we implemented a series of circuit breakers for each library that we depend on. Each circuit breaker controls the interaction between the API and that dependency. This image is a view of the dependency monitor that allows us to view the health and activity of each dependency. This dashboard is designed to give a real-time view of what is happening with these dependencies (over the last two minutes). We have other dashboards that provide insight into longer-term trends, day-over-day views, etc.
  • This is a view of a single circuit.
  • This circle represents the call volume and health of the dependency over the last 10 seconds. This circle is meant to be a visual indicator for health. The circle is green for healthy, yellow for borderline, and red for unhealthy. Moreover, the size of the circle represents the call volumes, where bigger circles mean more traffic.
  • The blue line represents the traffic trends over the last two minutes for this dependency.
  • The green number shows the number of successful calls to this dependency over the last two minutes.
  • The yellow number shows the number of latent calls into the dependency. These calls ultimately return successful responses, but slower than expected.
  • The blue number shows the number of calls that were handled by the short-circuited fallback mechanisms. That is, if the circuit gets tripped, the blue number will start to go up.
  • The orange number shows the number of calls that have timed out, resulting in fallback responses.
  • The purple number shows the number of calls that fail due to queuing issues, resulting in fallback responses.
  • The red number shows the number of exceptions, resulting in fallback responses.
  • The error rate is calculated from the total number of error and fallback responses divided by the total number calls handled.
  • If the error rate exceeds a certain number, the circuit to the fallback scenario is automatically opened. When it returns below that threshold, the circuit is closed again.
  • The dashboard also shows host and cluster information for the dependency.
  • As well as information about our SLAs.
  • So, going back to the engineering diagram…
  • If that same service fails today…
  • We simply disconnect from that service.
  • And replace it with an appropriate fallback. The fallback, ideally is a slightly degrade, but useful offering. If we cannot get that, however, we will quickly provide a 5xx response which will help the systems shed load rather than queue things up (which could eventually cause the system as a whole to tip over).
  • This will keep our customers happy, even if the experience may be slightly degraded. It is important to note that different dependency libraries have different fallback scenarios. And some are more resilient than others. But the overall sentiment here is accurate at a high level.
  • In addition to the migration to a distributed architecture, we also aggressively moved out of data centers…
  • And into the cloud.
  • Instead of spending in data centers, we spend out time in tools such as Asgard, created by Netflix staff, to help us manage our instance types and counts in AWS. Asgard is available in our open source repository at github.
  • Another feature afforded to us through AWS to help us scale is Autoscaling. This is the Netflix API request rates over a span of time. The red line represents a potential capacity needed in a data center to ensure that the spikes could be handled without spending a ton more than is needed for the really unlikely scenarios.
  • Through autoscaling, instead of buying new servers based on projected spikes in traffic and having systems administrators add them to the farm, the cloud can dynamically and automatically add and remove servers based on need.
  • To offset these limitations, we created Scryer (not yet open sourced, but in production at Netflix).
  • Instead of reacting to real-time metrics, like load average, to increase/decrease the instance count, we can look at historical patterns in our traffic to figure out what will be needed BEFORE it is needed. We believed we could write algorithms to predict the needs.
  • This is the result of the algorithms we created for the predictions. The prediction closely matches the actual traffic.
  • Going global has a different set of scaling challenges. AWS enables us to add instances in new regions that are closer to our customers.
  • To help us manage our traffic across regions, as well as within given regions, we created Zuul. Zuul is open source in our github repository.
  • Zuul does a variety of things for us. Zuul fronts our entire streaming application as well as a range of other services within our system.
  • Hystrix and other techniques throughout our engineering organization help keep things resilient. We also have an army of tools that introduce failures to the system which will help us identify problems before they become really big problems.
  • Hystrix and other techniques throughout our engineering organization help keep things resilient. We also have an army of tools that introduce failures to the system which will help us identify problems before they become really big problems.
  • The army is the Simian Army, which is a fleet of monkeys who are designed to do a variety of things, in an automated way, in our cloud implementation. Chaos Monkey, for example, periodically terminates AWS instances in production to see how the system as a whole will respond once that server disappears. Latency Monkey introduces latencies and errors into a system to see how it responds. The system is too complex to know how things will respond in various circumstances, so the monkeys expose that information to us in a variety of ways. The monkeys are also available in our open source github repository.
  • Again, the dependency chains in our system are quite complicated.
  • That is a lot of change in the system!
  • As a result, our philosophy is to act fast (ie. get code into production as quickly as possible), then react fast (ie. response to issues quickly as they arise).
  • Two such examples are canary deployments and what we call red/black deployments.
  • The canary deployments are comparable to canaries in coal mines. We have many servers in production running the current codebase. We will then introduce a single (or perhaps a few) new server(s) into production running new code. Monitoring the canary servers will show what the new code will look like in production.
  • If the canary encounters problems, it will register in any number of ways. The problems will be determined based on a comprehensive set of tools that will automatically perform health analysis on the canary.
  • The health of the canary is automated as well, comparing its metrics against the fleet of production servers.
  • If the canary encounters problems, it will register in any number of ways. The problems will be determined based on a comprehensive set of tools that will automatically perform health analysis on the canary.
  • If the canary shows errors, we pull it/them down, re-evaluate the new code, debug it, etc.
  • We will then repeat the process until the analysis of canary servers look good.
  • We will then repeat the process until the analysis of canary servers look good.
  • We also use Zuul to funnel varying degrees of traffic to the canaries to evaluate how much load the canary can take relative to the current production instances. If the RPS, for example, drops, the canary may fail the Zuul stress test.
  • If the new code looks good in the canary, we can then use a technique that we call red/black deployments to launch the code. Start with red, where production code is running. Fire up a new set of servers (black) equal to the count in red with the new code.
  • Then switch the pointer to have external requests point to the black servers. Sometimes, however, we may find an error in the black cluster that was not detected by the canary. For example, some issues can only be seen with full load.
  • Then switch the pointer to have external requests point to the black servers. Sometimes, however, we may find an error in the black cluster that was not detected by the canary. For example, some issues can only be seen with full load.
  • If a problem is encountered from the black servers, it is easy to rollback quickly by switching the pointer back to red. We will then re-evaluate the new code, debug it, etc.
  • Once we have debugged the code, we will put another canary up to evaluate the new changes in production.
  • And we will stress the canary again…
  • If the new code looks good in the canary, we can then bring up another set of servers with the new code.
  • Then we will switch production traffic to the new code.
  • If everything still looks good, we disable the red servers and the new code becomes the new red servers.
  • All of the open source components discussed here, as well as many others, can be found at the Netflix github repository.
  • Maintaining the Front Door to Netflix

    1. 1. Maintaining the Front Door to Netflix Ben Schmaus, @schmaus APIcon SF - 2014
    2. 2. Global Streaming Video for TV Shows and Movies
    3. 3. More than 48 Million Subscribers More than 40 Countries
    4. 4. Netflix Accounts for >34% of Peak Downstream Traffic in North America Netflix subscribers are watching more than 1 billion hours a month
    5. 5. Netflix Accounts for >6% of Peak Upstream Traffic in North America Netflix subscribers are watching more than 1 billion hours a month
    6. 6. Team Focus: Build the Best Global Streaming Product Three aspects of the Streaming Product: • Non-Member • Discovery • Streaming
    7. 7. Netflix API : Key Responsibilities • Broker data between services and Devices • Provide features and business logic • Maintain a resilient front-door • Scale the system • Maintain high velocity
    8. 8. Netflix API : Key Responsibilities • Broker data between services and Devices • Provide features and business logic • Maintain a resilient front-door • Scale the system • Maintain high velocity
    9. 9. APIs Do Lots of Things!
    10. 10. Data Gathering Data Formatting Data Delivery Security Authorization Authentication System Scaling Discoverability Data Consistency Translations Throttling Orchestration APIs Do Lots of Things! These are some of the many things APIs do.
    11. 11. Data Gathering Data Formatting Data Delivery Security Authorization Authentication System Scaling Discoverability Data Consistency Translations Throttling Orchestration APIs Do Lots of Things! These three are at the core. All others ultimately support them.
    12. 12. Definitions • Data Gathering – Retrieving the requested data from one or many local or remote data sources • Data Formatting – Preparing a structured payload to the requesting agent • Data Delivery – Delivering the structured payload to the requesting agent
    13. 13. Meanwhile… There are two players in APIs
    14. 14. API Provider
    15. 15. API Provider API Consumer
    16. 16. API Provider PROVIDES API Consumer CONSUMES Traditional API Interactions
    17. 17. API Provider PROVIDES EVERYTHING API Consumer CONSUMES Everything means, API Provider does: • Data Gathering • Data Formatting • Data Delivery • (among other things) Traditional API Interactions
    18. 18. Why do most API providers provide everything? • API design tends to be easier for teams closer to the source • Centralized API functions makes them easier to support • Many APIs have a large set of unknown and external developers
    19. 19. Why do most API providers provide everything? • API design tends to be easier for teams closer to the source • Centralized API functions makes them easier to support • Many APIs have a large set of unknown and external developers
    20. 20. Data Gathering Data Formatting Data Delivery API Consumer API Provider Separation of Concerns To be a better provider, the API should address the separation of concerns of the three core functions
    21. 21. Data Gathering Data Formatting Data Delivery API Consumer Don’t care how data is gathered, as long as it is gathered API Provider Care a lot about how the data is gathered Separation of Concerns
    22. 22. Data Gathering Data Formatting Data Delivery API Consumer Don’t care how data is gathered, as long as it is gathered Each consumer cares a lot about the format for that specific use API Provider Care a lot about how the data is gathered Only cares about the format to the extent it is easy to support Separation of Concerns
    23. 23. Data Gathering Data Formatting Data Delivery API Consumer Don’t care how data is gathered, as long as it is gathered Each consumer cares a lot about the format for that specific use Each consumer cares a lot about how payload is delivered API Provider Care a lot about how the data is gathered Only cares about the format to the extent it is easy to support Only cares about delivery method to the extent it is easy to support Separation of Concerns
    24. 24. Lsud vs. sskd
    25. 25. Should you consider alternatives to one-size-fits-all API model? Ingredients: • Small number of targeted API consumers is top priority • Close relationships between these API consumers and the API team • Increasing divergence of needs across the top priority API consumers • Strong desire by the API consumers for more optimized interactions with the API • High value proposition for the company providing the API to make these API consumers as effective as possible
    26. 26. • http://thenextweb.com/dd/2013/12/17/futur e-api-design-orchestration-layer/
    27. 27. Because of our separation of concerns, the Netflix API team is enabled to focus on different charters
    28. 28. Brokering Data to 1,000+ Device Types
    29. 29. Screen Real Estate
    30. 30. Controller
    31. 31. Technical Capabilities
    32. 32. One-Size-Fits-All API Request Request Request
    33. 33. Courtesy of South Florida Classical Review
    34. 34. Resource-Based API vs. Experience-Based API
    35. 35. Resource-Based Requests • /users/<id>/ratings/title • /users/<id>/queues • /users/<id>/queues/instant • /users/<id>/recommendations • /catalog/titles/movie • /catalog/titles/series • /catalog/people
    36. 36. REST API RECOMME NDATIONS MOVIE DATA SIMILAR MOVIES AUTH MEMBER DATA A/B TESTS START- UP RATINGS Network Border Network Border
    37. 37. RECOMME NDATIONS MOVIE DATA SIMILAR MOVIES AUTH MEMBER DATA A/B TESTS START- UP RATINGS OSFA API Network Border Network Border SERVER CODE CLIENT CODE
    38. 38. RECOMME NDATIONS MOVIE DATA SIMILAR MOVIES AUTH MEMBER DATA A/B TESTS START- UP RATINGS OSFA API Network Border Network Border DATA GATHERING, FORMATTING, AND DELIVERY USER INTERFACE RENDERING
    39. 39. Experience-Based Requests • /ps3/homescreen
    40. 40. JAVA API Network Border Network Border RECOMME NDATIONS MOVIE DATA SIMILAR MOVIES AUTH MEMBER DATA A/B TESTS START- UP RATINGS Groovy Layer
    41. 41. RECOMME NDATIONSA ZXSXX C CCC MOVIE DATA SIMILAR MOVIES AUTH MEMBER DATA A/B TESTS START- UP RATINGS JAVA API SERVER CODE CLIENT CODE CLIENT ADAPTER CODE (WRITTEN BY CLIENT TEAMS, DYNAMICALLY UPLOADED TO SERVER) Network Border Network Border
    42. 42. RECOMME NDATIONSA ZXSXX C CCC MOVIE DATA SIMILAR MOVIES AUTH MEMBER DATA A/B TESTS START- UP RATINGS JAVA API DATA GATHERING DATA FORMATTING AND DELIVERY USER INTERFACE RENDERING Network Border Network Border
    43. 43. Because we are no longer catering to LSUDs We can now focus on building a business on which all of Netflix can operate
    44. 44. Netflix API : Key Responsibilities • Broker data between services and Devices • Provide features and business logic • Maintain a resilient front-door • Scale the system • Maintain high velocity
    45. 45. The bigger the ship… the slower it turns
    46. 46. Distributed Architecture
    47. 47. 1000+ Device Types
    48. 48. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies Reviews A/B Test Engine Dozens of Dependencies
    49. 49. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    50. 50. Dependency Relationships
    51. 51. 2,000,000,000 Requests Per Day to the Netflix API
    52. 52. 30 Distinct Dependent Services for the Netflix API
    53. 53. ~600 Dependency jars Slurped into the Netflix API
    54. 54. 14,000,000,000 Netflix API Calls Per Day to those Dependent Services
    55. 55. 0 Dependent Services with 100% SLA
    56. 56. 99.99% = 99.7%30 0.3% of 2B = 6M failures per day 2+ Hours of Downtime Per Month
    57. 57. 99.99% = 99.7%30 0.3% of 2B = 6M failures per day 2+ Hours of Downtime Per Month
    58. 58. 99.9% = 97%30 3% of 2B = 60M failures per day 20+ Hours of Downtime Per Month
    59. 59. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    60. 60. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    61. 61. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    62. 62. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    63. 63. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    64. 64. Circuit Breaker Dashboard
    65. 65. Call Volume and Health / Last 10 Seconds
    66. 66. Call Volume / Last 2 Minutes
    67. 67. Successful Requests
    68. 68. Successful, But Slower Than Expected
    69. 69. Short-Circuited Requests, Delivering Fallbacks
    70. 70. Timeouts, Delivering Fallbacks
    71. 71. Thread Pool & Task Queue Full, Delivering Fallbacks
    72. 72. Exceptions, Delivering Fallbacks
    73. 73. Error Rate # + # + # + # / (# + # + # + # + #) = Error Rate
    74. 74. Status of Fallback Circuit
    75. 75. Requests per Second, Over Last 10 Seconds
    76. 76. SLA Information
    77. 77. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    78. 78. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    79. 79. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine
    80. 80. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine Fallback
    81. 81. Personaliz ation Engine User Info Movie Metadata Movie Ratings Similar Movies API Reviews A/B Test Engine Fallback
    82. 82. Netflix API : Key Responsibilities • Broker data between services and Devices • Provide features and business logic • Maintain a resilient front-door • Scale the system • Maintain high velocity
    83. 83. Scaling the Distributed System
    84. 84. AWS Cloud
    85. 85. Autoscaling
    86. 86. Autoscaling
    87. 87. Scryer : Predictive Auto Scaling Not yet…
    88. 88. Typical Traffic Patterns Over Five Days
    89. 89. Predicted RPS Compared to Actual RPS
    90. 90. More than 48 Million Subscribers More than 40 Countries
    91. 91. Zuul Gatekeeper for the Netflix Streaming Application
    92. 92. Zuul * • Multi-Region Resiliency • Insights • Stress Testing • Canary Testing • Dynamic Routing • Load Shedding • Security • Static Response Handling • Authentication * Most closely resembles an API proxy
    93. 93. All of these approaches are designed to prevent failures…
    94. 94. But sometimes the best way to prevent failures is to force them!
    95. 95. I randomly terminate instances in production to identify dormant failures. Chaos Monkey
    96. 96. Netflix API : Key Responsibilities • Broker data between services and Devices • Provide features and business logic • Maintain a resilient front-door • Scale the system • Maintain high velocity
    97. 97. Dependency Relationships
    98. 98. Testing Philosophy: Act Fast, React Fast
    99. 99. That Doesn’t Mean We Don’t Test
    100. 100. Automated Delivery Pipeline
    101. 101. Cloud-Based Deployment Techniques
    102. 102. Current Code In Production API Requests from the Internet
    103. 103. Single Canary Instance To Test New Code with Production Traffic (around 1% or less of traffic) Current Code In Production API Requests from the Internet
    104. 104. Canary Analysis Automation
    105. 105. Single Canary Instance To Test New Code with Production Traffic (around 1% or less of traffic) Current Code In Production API Requests from the Internet Error!
    106. 106. Current Code In Production API Requests from the Internet
    107. 107. Current Code In Production API Requests from the Internet
    108. 108. Current Code In Production API Requests from the Internet Perfect!
    109. 109. Stress Test with Zuul
    110. 110. Current Code In Production API Requests from the Internet New Code Getting Prepared for Production
    111. 111. Current Code In Production API Requests from the Internet New Code Getting Prepared for Production
    112. 112. Error! Current Code In Production API Requests from the Internet New Code Getting Prepared for Production
    113. 113. Current Code In Production API Requests from the Internet New Code Getting Prepared for Production
    114. 114. Current Code In Production API Requests from the Internet Perfect!
    115. 115. Stress Test with Zuul
    116. 116. Current Code In Production API Requests from the Internet New Code Getting Prepared for Production
    117. 117. Current Code In Production API Requests from the Internet New Code Getting Prepared for Production
    118. 118. API Requests from the Internet New Code Getting Prepared for Production
    119. 119. https://www.github.com/Netflix
    120. 120. Maintaining the Front Door to Netflix Ben Schmaus, @schmaus APIcon SF - 2014

    ×