May 18, 2024

David J. Berg, Romain Cledat, Kayla Seeley, Shashank Srikanth, Chaoying Wang, Darin Yu

Netflix makes use of information science and machine studying throughout all aspects of the corporate, powering a variety of enterprise purposes from our inner infrastructure and content material demand modeling to media understanding. The Machine Studying Platform (MLP) crew at Netflix supplies a complete ecosystem of instruments round Metaflow, an open supply machine studying infrastructure framework we began, to empower information scientists and machine studying practitioners to construct and handle a wide range of ML methods.

Since its inception, Metaflow has been designed to offer a human-friendly API for constructing information and ML (and at the moment AI) purposes and deploying them in our manufacturing infrastructure frictionlessly. Whereas human-friendly APIs are pleasant, it’s actually the integrations to our manufacturing methods that give Metaflow its superpowers. With out these integrations, tasks could be caught on the prototyping stage, or they must be maintained as outliers exterior the methods maintained by our engineering groups, incurring unsustainable operational overhead.

Given the very various set of ML and AI use circumstances we help — at the moment we have now lots of of Metaflow tasks deployed internally — we don’t count on all tasks to comply with the identical path from prototype to manufacturing. As a substitute, we offer a strong foundational layer with integrations to our company-wide information, compute, and orchestration platform, in addition to varied paths to deploy purposes to manufacturing easily. On high of this, groups have constructed their very own domain-specific libraries to help their particular use circumstances and wishes.

On this article, we cowl a number of key integrations that we offer for varied layers of the Metaflow stack at Netflix, as illustrated above. We will even showcase real-life ML tasks that depend on them, to provide an concept of the breadth of tasks we help. Notice that each one tasks leverage a number of integrations, however we spotlight them within the context of the combination that they use most prominently. Importantly, all of the use circumstances have been engineered by practitioners themselves.

These integrations are carried out by Metaflow’s extension mechanism which is publicly out there however topic to alter, and therefore not part of Metaflow’s secure API but. If you’re inquisitive about implementing your individual extensions, get in contact with us on the Metaflow community Slack.

Let’s go over the stack layer by layer, beginning with essentially the most foundational integrations.

Our fundamental information lake is hosted on S3, organized as Apache Iceberg tables. For ETL and different heavy lifting of information, we primarily depend on Apache Spark. Along with Spark, we need to help last-mile information processing in Python, addressing use circumstances equivalent to function transformations, batch inference, and coaching. Often, these use circumstances contain terabytes of information, so we have now to concentrate to efficiency.

To allow quick, scalable, and sturdy entry to the Netflix information warehouse, we have now developed a Quick Knowledge library for Metaflow, which leverages high-performance parts from the Python information ecosystem:

As depicted within the diagram, the Quick Knowledge library consists of two fundamental interfaces:

  • The Desk object is chargeable for interacting with the Netflix information warehouse which incorporates parsing Iceberg (or legacy Hive) desk metadata, resolving partitions and Parquet recordsdata for studying. Lately, we added help for the write path, so tables might be up to date as properly utilizing the library.
  • As soon as we have now found the Parquet recordsdata to be processed, MetaflowDataFrame takes over: it downloads information utilizing Metaflow’s high-throughput S3 consumer on to the method’ reminiscence, which often outperforms reading of local files.

We use Apache Arrow to decode Parquet and to host an in-memory illustration of information. The person can select essentially the most appropriate device for manipulating information, equivalent to Pandas or Polars to make use of a dataframe API, or one in every of our inner C++ libraries for varied high-performance operations. Because of Arrow, information might be accessed by these libraries in a zero-copy trend.

We additionally take note of dependency points: (Py)Arrow is a dependency of many ML and information libraries, so we don’t need our customized C++ extensions to rely upon a particular model of Arrow, which might simply result in unresolvable dependency graphs. As a substitute, within the type of nanoarrow, our Quick Knowledge library solely depends on the stable Arrow C data interface, producing a hermetically sealed library with no exterior dependencies.

Instance use case: Content material Data Graph

Our data graph of the leisure world encodes relationships between titles, actors and different attributes of a movie or collection, supporting all points of enterprise at Netflix.

A key problem in making a data graph is entity decision. There could also be many alternative representations of barely completely different or conflicting details about a title which should be resolved. That is sometimes accomplished by a pairwise matching process for every entity which turns into non-trivial to do at scale.

This undertaking leverages Quick Knowledge and horizontal scaling with Metaflow’s foreach construct to load massive quantities of title data — roughly a billion pairs — saved within the Netflix Knowledge Warehouse, so the pairs might be matched in parallel throughout many Metaflow duties.

We use metaflow.Desk to resolve all enter shards that are distributed to Metaflow duties that are chargeable for processing terabytes of information collectively. Every process hundreds the information utilizing metaflow.MetaflowDataFrame, performs matching utilizing Pandas, and populates a corresponding shard in an output Desk. Lastly, when all matching is finished and information is written the brand new desk is dedicated so it may be learn by different jobs.

Whereas open-source customers of Metaflow depend on AWS Batch or Kubernetes as the compute backend, we depend on our centralized compute-platform, Titus. Beneath the hood, Titus is powered by Kubernetes, but it surely supplies a thick layer of enhancements over off-the-shelf Kubernetes, to make it extra observable, safe, scalable, and cost-efficient.

By focusing on @titus, Metaflow duties profit from these battle-hardened options out of the field, with no in-depth technical data or engineering required from the ML engineers or information scientist finish. Nonetheless, in an effort to profit from scalable compute, we have to assist the developer to package deal and rehydrate the entire execution surroundings of a undertaking in a distant pod in a reproducible method (ideally rapidly). Particularly, we don’t need to ask builders to handle Docker pictures of their very own manually, which rapidly leads to extra issues than it solves.

This is the reason Metaflow provides support for dependency management out of the field. Initially, we supported solely @conda, however based mostly on our work on Portable Execution Environments, open-source Metaflow gained support for @pypi a number of months in the past as properly.

Instance use case: Constructing mannequin explainers

Right here’s an enchanting instance of the usefulness of transportable execution environments. For a lot of of our purposes, mannequin explainability issues. Stakeholders like to grasp why fashions produce a sure output and why their habits adjustments over time.

There are a number of methods to offer explainability to fashions however a method is to coach an explainer mannequin based mostly on every skilled mannequin. With out going into the main points of how that is accomplished precisely, suffice to say that Netflix trains a number of fashions, so we have to practice a number of explainers too.

Because of Metaflow, we are able to permit every utility to decide on the very best modeling strategy for his or her use circumstances. Correspondingly, every utility brings its personal bespoke set of dependencies. Coaching an explainer mannequin subsequently requires:

  1. Entry to the unique mannequin and its coaching surroundings, and
  2. Dependencies particular to constructing the explainer mannequin.

This poses an attention-grabbing problem in dependency administration: we’d like a higher-order coaching system, “Explainer circulation” within the determine under, which is ready to take a full execution surroundings of one other coaching system as an enter and produce a mannequin based mostly on it.

Explainer circulation is event-triggered by an upstream circulation, such Mannequin A, B, C flows within the illustration. The build_environment step makes use of the metaflow surroundings command offered by our portable environments, to construct an surroundings that features each the necessities of the enter mannequin in addition to these wanted to construct the explainer mannequin itself.

The constructed surroundings is given a singular title that depends upon the run identifier (to offer uniqueness) in addition to the mannequin kind. Given this surroundings, the train_explainer step is then in a position to consult with this uniquely named surroundings and function in an surroundings that may each entry the enter mannequin in addition to practice the explainer mannequin. Notice that, in contrast to in typical flows utilizing vanilla @conda or @pypi, the transportable environments extension permits customers to additionally fetch these environments immediately at execution time versus at deploy time which subsequently permits customers to, as on this case, resolve the surroundings proper earlier than utilizing it within the subsequent step.

If information is the gas of ML and the compute layer is the muscle, then the nerves should be the orchestration layer. We’ve talked concerning the significance of a production-grade workflow orchestrator within the context of Metaflow once we launched help for AWS Step Features years in the past. Since then, open-source Metaflow has gained help for Argo Workflows, a Kubernetes-native orchestrator, in addition to support for Airflow which continues to be extensively utilized by information engineering groups.

Internally, we use a manufacturing workflow orchestrator referred to as Maestro. The Maestro put up shares particulars about how the system helps scalability, high-availability, and value, which give the spine for all of our Metaflow tasks in manufacturing.

A vastly vital element that always goes ignored is event-triggering: it permits a crew to combine their Metaflow flows to surrounding methods upstream (e.g. ETL workflows), in addition to downstream (e.g. flows managed by different groups), utilizing a protocol shared by the entire group, as exemplified by the instance use case under.

Instance use case: Content material resolution making

One of the crucial business-critical methods operating on Metaflow helps our content material resolution making, that’s, the query of what content material Netflix ought to convey to the service. We help a large scale of over 260M subscribers spanning over 190 international locations representing vastly various cultures and tastes, all of whom we need to delight with our content material slate. Reflecting the breadth and depth of the problem, the methods and fashions specializing in the query have grown to be very refined.

We strategy the query from a number of angles however we have now a core set of information pipelines and fashions that present a basis for resolution making. For example the complexity of simply the core parts, take into account this high-level diagram:

On this diagram, grey containers characterize integrations to accomplice groups downstream and upstream, inexperienced containers are varied ETL pipelines, and blue containers are Metaflow flows. These containers encapsulate lots of of superior fashions and complex enterprise logic, dealing with large quantities of information day by day.

Regardless of its complexity, the system is managed by a comparatively small crew of engineers and information scientists autonomously. That is made attainable by a number of key options of Metaflow:

The crew has additionally developed their very own domain-specific libraries and configuration administration instruments, which assist them enhance and function the system.

To provide enterprise worth, all our Metaflow tasks are deployed to work with different manufacturing methods. In lots of circumstances, the combination is likely to be through shared tables in our information warehouse. In different circumstances, it’s extra handy to share the outcomes through a low-latency API.

Notably, not all API-based deployments require real-time analysis, which we cowl within the part under. We’ve plenty of business-critical purposes the place some or all predictions might be precomputed, guaranteeing the bottom attainable latency and operationally easy excessive availability on the world scale.

We’ve developed an formally supported sample to cowl such use circumstances. Whereas the system depends on our inner caching infrastructure, you would comply with the identical sample utilizing providers like Amazon ElasticCache or DynamoDB.

Instance use case: Content material efficiency visualization

The historic efficiency of titles is utilized by resolution makers to grasp and enhance the movie and collection catalog. Efficiency metrics might be complicated and are sometimes greatest understood by people with visualizations that break down the metrics throughout parameters of curiosity interactively. Content material resolution makers are outfitted with self-serve visualizations by a real-time net utility constructed with metaflow.Cache, which is accessed by an API supplied with metaflow.Internet hosting.

A day by day scheduled Metaflow job computes combination portions of curiosity in parallel. The job writes a big quantity of outcomes to a web based key-value retailer utilizing metaflow.Cache. A Streamlit app homes the visualization software program and information aggregation logic. Customers can dynamically change parameters of the visualization utility and in real-time a message is shipped to a easy Metaflow hosting service which seems to be up values within the cache, performs computation, and returns the outcomes as a JSON blob to the Streamlit utility.

For deployments that require an API and real-time analysis, we offer an built-in mannequin internet hosting service, Metaflow Internet hosting. Though particulars have developed quite a bit, this old talk still gives a good overview of the service.

Metaflow Internet hosting is particularly geared in direction of internet hosting artifacts or fashions produced in Metaflow. This supplies a simple to make use of interface on high of Netflix’s current microservice infrastructure, permitting information scientists to rapidly transfer their work from experimentation to a manufacturing grade net service that may be consumed over a HTTP REST API with minimal overhead.

Its key advantages embody:

  • Easy decorator syntax to create RESTFull endpoints.
  • The back-end auto-scales the variety of cases used to again your service based mostly on visitors.
  • The back-end will scale-to-zero if no requests are made to it after a specified period of time thereby saving price significantly in case your service requires GPUs to successfully produce a response.
  • Request logging, alerts, monitoring and tracing hooks to Netflix infrastructure

Contemplate the service much like managed mannequin internet hosting providers like AWS Sagemaker Model Hosting, however tightly built-in with our microservice infrastructure.

Instance use case: Media

We’ve a protracted historical past of utilizing machine studying to course of media property, as an example, to personalize paintings and to assist our creatives create promotional content material effectively. Processing massive quantities of media property is technically non-trivial and computationally costly, so through the years, we have now developed loads of specialised infrastructure devoted for this goal usually, and infrastructure supporting media ML use circumstances specifically.

To display the advantages of Metaflow Internet hosting that gives a general-purpose API layer supporting each synchronous and asynchronous queries, take into account this use case involving Amber, our function retailer for media.

Whereas Amber is a function retailer, precomputing and storing all media options upfront could be infeasible. As a substitute, we compute and cache options in an on-demand foundation, as depicted under:

When a service requests a function from Amber, it computes the function dependency graph after which sends a number of asynchronous requests to Metaflow Internet hosting, which locations the requests in a queue, ultimately triggering function computations when compute sources turn out to be out there. Metaflow Internet hosting caches the response, so Amber can fetch it after some time. We might have constructed a devoted microservice only for this use case, however due to the flexibleness of Metaflow Internet hosting, we have been in a position to ship the function sooner with no extra operational burden.

Our urge for food to use ML in various use circumstances is just rising, so our Metaflow platform will hold increasing its footprint correspondingly and proceed to offer pleasant integrations to methods constructed by different groups at Netlfix. As an example, we have now plans to work on enhancements within the versioning layer, which wasn’t coated by this text, by giving extra choices for artifact and mannequin administration.

We additionally plan on constructing extra integrations with different methods which can be being developed by sister groups at Netflix. For instance, Metaflow Internet hosting fashions are presently not properly built-in into mannequin logging services — we plan on engaged on bettering this to make fashions developed with Metaflow extra built-in with the suggestions loop crucial in coaching new fashions. We hope to do that in a pluggable method that might permit different customers to combine with their very own logging methods.

Moreover we need to provide extra methods Metaflow artifacts and fashions might be built-in into non-Metaflow environments and purposes, e.g. JVM based mostly edge service, in order that Python-based information scientists can contribute to non-Python engineering methods simply. This is able to permit us to raised bridge the hole between the fast iteration that Metaflow supplies (in Python) with the necessities and constraints imposed by the infrastructure serving Netflix member dealing with requests.

If you’re constructing business-critical ML or AI methods in your group, join the Metaflow Slack community! We’re comfortable to share experiences, reply any questions, and welcome you to contribute to Metaflow.


Because of Wenbing Bai, Jan Florjanczyk, Michael Li, Aliki Mavromoustaki, and Sejal Rai for assist with use circumstances and figures. Because of our OSS contributors for making Metaflow a greater product.