Scalable Annotation Service — Marken | Netflix TechBlog

At Netflix, we have now a whole lot of micro companies every with its personal knowledge fashions or entities. For instance, we have now a service that shops a film entity’s metadata or a service that shops metadata about pictures. All of those companies at a later level need to annotate their objects or entities. Our workforce, Asset Administration Platform, determined to create a generic service referred to as Marken which permits any microservice at Netflix to annotate their entity.


Generally individuals describe annotations as tags however that could be a restricted definition. In Marken, an annotation is a bit of metadata which could be hooked up to an object from any area. There are numerous totally different sorts of annotations our shopper functions need to generate. A easy annotation, like beneath, would describe {that a} specific film has violence.

  • Film Entity with id 1234 has violence.

However there are extra fascinating circumstances the place customers need to retailer temporal (time-based) knowledge or spatial knowledge. In Pic 1 beneath, we have now an instance of an software which is utilized by editors to assessment their work. They need to change the colour of gloves to wealthy black so they need to have the ability to mark up that space, on this case utilizing a blue circle, and retailer a remark for it. It is a typical use case for a artistic assessment software.

An instance for storing each time and area primarily based knowledge can be an ML algorithm that may determine characters in a body and desires to retailer the next for a video

  • In a specific body (time)
  • In some space in picture (area)
  • A personality title (annotation knowledge)
Pic 1 : Editors requesting adjustments by drawing shapes just like the blue circle proven above.

Objectives for Marken

We needed to create an annotation service which could have the next targets.

  • Permits to annotate any entity. Groups ought to have the ability to outline their knowledge mannequin for annotation.
  • Annotations could be versioned.
  • The service ought to have the ability to serve real-time, aka UI, functions so CRUD and search operations must be achieved with low latency.
  • All knowledge must be additionally accessible for offline analytics in Hive/Iceberg.


Because the annotation service can be utilized by anybody at Netflix we had a have to help totally different knowledge fashions for the annotation object. An information mannequin in Marken could be described utilizing schema — similar to how we create schemas for database tables and so forth.

Our workforce, Asset Administration Platform, owns a special service that has a json primarily based DSL to explain the schema of a media asset. We prolonged this service to additionally describe the schema of an annotation object.

"sort": "BOUNDING_BOX", ❶
"model": 0, ❷
"description": "Schema describing a bounding field",
"properties": ❸
"sort": "bounding_box",
"necessary": true
"sort": "time_range",
"necessary": true

Within the above instance, the appliance needs to characterize in a video an oblong space which spans a variety of time.

  1. Schema’s title is BOUNDING_BOX
  2. Schemas can have variations. This enables customers to make add/take away properties of their knowledge mannequin. We don’t enable incompatible adjustments, for instance, customers cannot change the information sort of a property.
  3. The info saved is represented within the “properties” part. On this case, there are two properties
  4. boundingBox, with sort “bounding_box”. That is principally an oblong space.
  5. boxTimeRange, with sort “time_range”. This enables us to specify begin and finish time for this annotation.

Geometry Objects

To characterize spatial knowledge in an annotation we used the Well Known Text (WKT) format. We help following objects

  • Level
  • Line
  • MultiLine
  • BoundingBox
  • LinearRing

Our mannequin is extensible permitting us to simply add extra geometry objects as wanted.

Temporal Objects

A number of functions have a requirement to retailer annotations for movies which have time in it. We enable functions to retailer time as body numbers or nanoseconds.

To retailer knowledge in frames purchasers should additionally retailer frames per second. We name this a SampleData with following elements:

  • sampleNumber aka body quantity
  • sampleNumerator
  • sampleDenominator

Annotation Object

Identical to schema, an annotation object can be represented in JSON. Right here is an instance of annotation for BOUNDING_BOX which we mentioned above.

"annotationId": ❶
"id": "188c5b05-e648-4707-bf85-dada805b8f87",
"model": "0"
"associatedId": ❷
"entityType": "MOVIE_ID",
"id": "1234"
"annotationType": "ANNOTATION_BOUNDINGBOX", ❸
"annotationTypeVersion": 1,
"metadata": ❹
"fileId": "identityOfSomeFile",
"x": 20,
"y": 30
"x": 40,
"y": 60

"startTimeInNanoSec": 566280000000,
"endTimeInNanoSec": 567680000000

  1. The primary part is the distinctive id of this annotation. An annotation is an immutable object so the id of the annotation all the time features a model. At any time when somebody updates this annotation we mechanically increment its model.
  2. An annotation should be related to some entity which belongs to some microservice. On this case, this annotation was created for a film with id “1234”
  3. We then specify the schema sort of the annotation. On this case it’s BOUNDING_BOX.
  4. Precise knowledge is saved within the metadata part of json. Like we mentioned above there’s a bounding field and time vary in nanoseconds.

Base schemas

Identical to in Object Oriented Programming, our schema service permits schemas to be inherited from one another. This enables our purchasers to create an “is-a-type-of” relationship between schemas. Not like Java, we help a number of inheritance as properly.

We’ve got a number of ML algorithms which scan Netflix media property (pictures and movies) and create very fascinating knowledge for instance figuring out characters in frames or figuring out match cuts. This knowledge is then saved as annotations in our service.

As a platform service we created a set of base schemas to ease creating schemas for various ML algorithms. One base schema (TEMPORAL_SPATIAL_BASE) has the next elective properties. This base schema can be utilized by any derived schema and never restricted to ML algorithms.

  • Temporal (time associated knowledge)
  • Spatial (geometry knowledge)

And one other one BASE_ALGORITHM_ANNOTATION which has the next elective properties which is often utilized by ML algorithms.

  • label (String)
  • confidenceScore (double) — denotes the boldness of the generated knowledge from the algorithm.
  • algorithmVersion (String) — model of the ML algorithm.

By utilizing a number of inheritance, a typical ML algorithm schema derives from each TEMPORAL_SPATIAL_BASE and BASE_ALGORITHM_ANNOTATION schemas.

"model": 0,
"description": "Base Schema for Algorithm primarily based Annotations",
"sort": "decimal",
"necessary": false,
"description": "Confidence Rating",
"sort": "string",
"necessary": false,
"description": "Annotation Tag",
"sort": "string",
"description": "Algorithm Model"


Given the targets of the service we needed to maintain following in thoughts.

  • Our service will probably be utilized by plenty of inner UI functions therefore the latency for CRUD and search operations should be low.
  • Apart from functions we could have ML algorithm knowledge saved. A few of this knowledge could be on the body stage for movies. So the quantity of knowledge saved could be massive. The databases we choose ought to have the ability to scale horizontally.
  • We additionally anticipated that the service could have excessive RPS.

Another targets got here from search necessities.

  • Capacity to look the temporal and spatial knowledge.
  • Capacity to look with totally different related and extra related Ids as described in our Annotation Object knowledge mannequin.
  • Full textual content searches on many alternative fields within the Annotation Object
  • Stem search help

As time progressed the necessities for search solely elevated and we are going to focus on these necessities intimately in a special part.

Given the necessities and the experience in our workforce we determined to decide on Cassandra because the supply of reality for storing annotations. For supporting totally different search necessities we selected ElasticSearch. Apart from to help varied options we have now bunch of inner auxiliary companies for eg. zookeeper service, internationalization service and so forth.

Marken structure

Above image represents the block diagram of the structure for our service. On the left we present knowledge pipelines that are created by a number of of our shopper groups to mechanically ingest new knowledge into our service. An important of such an information pipeline is created by the Machine Studying workforce.

One of many key initiatives at Netflix, Media Search Platform, now makes use of Marken to retailer annotations and carry out varied searches defined beneath. Our structure makes it doable to simply onboard and ingest knowledge from Media algorithms. This knowledge is utilized by varied groups for eg. creators of promotional media (aka trailers, banner pictures) to enhance their workflows.


Success of Annotation Service (knowledge labels) relies on the efficient search of these labels with out understanding a lot of enter algorithms particulars. As talked about above, we use the bottom schemas for each new annotation sort (relying on the algorithm) listed into the service. This helps our purchasers to look throughout the totally different annotation varieties persistently. Annotations could be searched both by merely knowledge labels or with extra added filters like film id.

We’ve got outlined a customized question DSL to help looking, sorting and grouping of the annotation outcomes. Various kinds of search queries are supported utilizing the Elasticsearch as a backend search engine.

  • Full Textual content Search — Purchasers could not know the precise labels created by the ML algorithms. For example, the label could be ‘bathe curtain’. With full textual content search, purchasers can discover the annotation by looking utilizing label ‘curtain’ . We additionally help fuzzy search on the label values. For instance, if the purchasers need to search ‘curtain’ however they wrongly typed ‘curtian` — annotation with the ‘curtain’ label will probably be returned.
  • Stem Search — With international Netflix content material supported in numerous languages, our purchasers have the requirement to help stem seek for totally different languages. Marken service incorporates subtitles for a full catalog of titles in Netflix which could be in many alternative languages. For example for stem search , `clothes` and `garments` could be stemmed to the identical root phrase `material`. We use ElasticSearch to help stem seek for 34 totally different languages.
  • Temporal Annotations Search — Annotations for movies are extra related whether it is outlined together with the temporal (time vary with begin and finish time) info. Time vary inside video can be mapped to the body numbers. We help labels seek for the temporal annotations inside the offered time vary/body quantity additionally.
  • Spatial Annotation Search — Annotations for video or picture can even embody the spatial info. For instance a bounding field which defines the placement of the labeled object within the annotation.
  • Temporal and Spatial Search — Annotation for video can have each time vary and spatial coordinates. Therefore, we help queries which may search annotations inside the offered time vary and spatial coordinates vary.
  • Semantics Search — Annotations could be searched after understanding the intent of the person offered question. Any such search gives outcomes primarily based on the conceptually related matches to the textual content within the question, in contrast to the normal tag primarily based search which is anticipated to be actual key phrase matches with the annotation labels. ML algorithms additionally ingest annotations with vectors as an alternative of precise labels to help this sort of search. Person offered textual content is transformed right into a vector utilizing the identical ML mannequin, after which search is carried out with the transformed text-to-vector to search out the closest vectors with the searched vector. Based mostly on the purchasers suggestions, such searches present extra related outcomes and don’t return empty leads to case there aren’t any annotations which precisely match to the person offered question labels. We help semantic search utilizing Open Distro for ElasticSearch . We’ll cowl extra particulars on Semantic Search help in a future weblog article.
Semantic search
  • Vary Intersection — We lately began supporting the vary intersection queries throughout a number of annotation varieties for a selected title in the true time. This enables the purchasers to look with a number of knowledge labels (resulted from totally different algorithms so they’re totally different annotation varieties) inside video particular time vary or the whole video, and get the listing of time ranges or frames the place the offered set of knowledge labels are current. A typical instance of this question is to search out the `James within the indoor shot consuming wine`. For such queries, the question processor finds the outcomes of each knowledge labels (James, Indoor shot) and vector search (consuming wine); after which finds the intersection of ensuing frames in-memory.

Search Latency

Our shopper functions are studio UI functions in order that they count on low latency for the search queries. As highlighted above, we help such queries utilizing Elasticsearch. To maintain the latency low, we have now to ensure that all of the annotation indices are balanced, and hotspot isn’t created with any algorithm backfill knowledge ingestion for the older motion pictures. We adopted the rollover indices technique to keep away from such hotspots (as described in our blog for asset administration software) within the cluster which may trigger spikes within the cpu utilization and decelerate the question response. Search latency for the generic textual content queries are in milliseconds. Semantic search queries have comparatively greater latency than generic textual content searches. Following graph reveals the common search latency for generic search and semantic search (together with KNN and ANN search) latencies.

Common search latency
Semantic search latency


One of many key challenges whereas designing the annotation service is to deal with the scaling necessities with the rising Netflix film catalog and ML algorithms. Video content material evaluation performs a vital position within the utilization of the content material throughout the studio functions within the film manufacturing or promotion. We count on the algorithm varieties to develop broadly within the coming years. With the rising variety of annotations and its utilization throughout the studio functions, prioritizing scalability turns into important.

Information ingestions from the ML knowledge pipelines are typically in bulk particularly when a brand new algorithm is designed and annotations are generated for the total catalog. We’ve got arrange a special stack (fleet of cases) to manage the information ingestion stream and therefore present constant search latency to our shoppers. On this stack, we’re controlling the write throughput to our backend databases utilizing Java threadpool configurations.

Cassandra and Elasticsearch backend databases help horizontal scaling of the service with rising knowledge dimension and queries. We began with a 12 nodes cassandra cluster, and scaled as much as 24 nodes to help present knowledge dimension. This 12 months, annotations are added roughly for the Netflix full catalog. Some titles have greater than 3M annotations (most of them are associated to subtitles). At present the service has round 1.9 billion annotations with knowledge dimension of two.6TB.


Annotations could be searched in bulk throughout a number of annotation varieties to construct knowledge details for a title or throughout a number of titles. For such use circumstances, we persist all of the annotation knowledge in iceberg tables in order that annotations could be queried in bulk with totally different dimensions with out impacting the true time functions CRUD operations latency.

One of many frequent use circumstances is when the media algorithm groups learn subtitle knowledge in numerous languages (annotations containing subtitles on a per body foundation) in bulk in order that they’ll refine the ML fashions they’ve created.

Future work

There may be plenty of fascinating future work on this space.

  1. Our knowledge footprint retains rising with time. A number of instances we have now knowledge from algorithms that are revised and annotations associated to the brand new model are extra correct and in-use. So we have to do cleanups for giant quantities of knowledge with out affecting the service.
  2. Intersection queries over a big scale of knowledge and returning outcomes with low latency is an space the place we need to make investments extra time.


Burak Bacioglu and different members of the Asset Administration Platform contributed within the design and growth of Marken.