Knowledge ingestion pipeline with Operation Administration (Marken)

At Netflix, to advertise and advocate the content material to customers in the absolute best manner there are various Media Algorithm groups which work hand in hand with content material creators and editors. A number of of those algorithms goal to enhance totally different handbook workflows in order that we present the customized promotional picture, trailer or the present to the consumer.

These media centered machine studying algorithms in addition to different groups generate plenty of information from the media recordsdata, which we described in our earlier weblog, are saved as annotations in Marken. We designed a novel idea known as Annotation Operations which permits groups to create information pipelines and simply write annotations with out worrying about entry patterns of their information from totally different purposes.

Annotation Operations

Lets choose an instance use case of figuring out objects (like bushes, vehicles and many others.) in a video file. As described within the above image

  • Through the first run of the algorithm it recognized 500 objects in a selected Video file. These 500 objects had been saved as annotations of a particular schema sort, let’s say Objects, in Marken.
  • The Algorithm crew improved their algorithm. Now once we re-ran the algorithm on the identical video file it created 600 annotations of schema sort Objects and saved them in our service.

Discover that we can not replace the annotations from earlier runs as a result of we don’t know what number of annotations a brand new algorithm run will consequence into. Additionally it is very costly for us to maintain monitor of which annotation must be up to date.

The purpose is that when the buyer comes and searches for annotations of sort Objects for the given video file then the next ought to occur.

  • Earlier than Algo run 1, in the event that they search they need to not discover something.
  • After the completion of Algo run 1, the question ought to discover the primary set of 500 annotations.
  • Through the time when Algo run 2 was creating the set of 600 annotations, purchasers search ought to nonetheless return the older 500 annotations.
  • When the entire 600 annotations are efficiently created, they need to exchange the older set of 500.
  • So now when purchasers search annotations for Objects then they need to get 600 annotations.

Does this remind you of one thing? This appears very comparable (not precisely similar) to a distributed transaction.

Usually, an algorithm run can have 2k-5k annotations. There are a lot of naive options doable for this downside for instance:

  • Write totally different runs in numerous databases. That is clearly very costly.
  • Write algo runs into recordsdata. However we can not search or current low latency retrievals from recordsdata
  • And many others.

As a substitute our problem was to implement this function on prime of Cassandra and ElasticSearch databases as a result of that’s what Marken makes use of. The answer which we current on this weblog isn’t restricted to annotations and can be utilized for some other area which makes use of ES and Cassandra as properly.

Marken’s structure diagram is as follows. We refer the reader to our earlier weblog article for particulars. We use Cassandra as a supply of fact the place we retailer the annotations whereas we index annotations in ElasticSearch to offer wealthy search functionalities.

Marken Structure

Our purpose was to assist groups at Netflix to create information pipelines with out fascinated by how that information is on the market to the readers or the shopper groups. Equally, shopper groups don’t have to fret about when or how the info is written. That is what we name decoupling producer flows from purchasers of the info.

Lifecycle of a film goes by plenty of inventive phases. We now have many momentary recordsdata that are delivered earlier than we get to the ultimate file of the film. Equally, a film has many alternative languages and every of these languages can have totally different recordsdata delivered. Groups typically wish to run algorithms and create annotations utilizing all these media recordsdata.

Since algorithms could be run on a unique permutations of how the media recordsdata are created and delivered we are able to simplify an algorithm run as follows

  • Annotation Schema Kind — identifies the schema for the annotation generated by the Algorithm.
  • Annotation Schema Model — identifies the schema model of the annotation generated by the Algorithm.
  • PivotId — a novel string identifier which identifies the file or technique which is used to generate the annotations. This might be the SHA hash of the file or just the film Identifier quantity.

Given above we are able to describe the info mannequin for an annotation operation as follows.

"annotationOperationKeys": [

"annotationType": "string", ❶
"annotationTypeVersion": “integer”,
"pivotId": "string",
"operationNumber": “integer” ❷

"id": "UUID",
"operationStatus": "STARTED", ❸
"isActive": true ❹
  1. We already defined AnnotationType, AnnotationTypeVersion and PivotId above.
  2. OperationNumber is an auto incremented quantity for every new operation.
  3. OperationStatus — An operation goes by three phases, Began, Completed and Canceled.
  4. IsActive — Whether or not an operation and its related annotations are energetic and searchable.

As you’ll be able to see from the info mannequin that the producer of an annotation has to decide on an AnnotationOperationKey which lets them outline how they need UPSERT annotations in an AnnotationOperation. Inside, AnnotationOperationKey the essential area is pivotId and the way it’s generated.

Our supply of fact for all objects in Marken in Cassandra. To retailer Annotation Operations we’ve the next major tables.

  • AnnotationOperationById — It shops the AnnotationOperations
  • AnnotationIdByAnnotationOperationId — it shops the Ids of all annotations in an operation.

Since Cassandra is NoSql, we’ve extra tables which assist us create reverse indices and run admin jobs in order that we are able to scan all annotation operations at any time when there’s a want.

Every annotation in Marken can also be listed in ElasticSearch for powering numerous searches. To file the connection between annotation and operation we additionally index two fields

  • annotationOperationId — The ID of the operation to which this annotation belongs
  • isAnnotationOperationActive — Whether or not the operation is in an ACTIVE state.

We offer three APIs to our customers. In following sections we describe the APIs and the state administration achieved inside the APIs.


When this API known as we retailer the operation with its OperationKey (tuple of annotationType, annotationType Model and pivotId) in our database. This new operation is marked to be in STARTED state. We retailer all OperationIDs that are in STARTED state in a distributed cache (EVCache) for quick entry throughout searches.



Customers name this API to upsert the annotations in an Operation. They move annotations together with the OperationID. We retailer the annotations and in addition file the connection between the annotation IDs and the Operation ID in Cassandra. Throughout this section operations are in isAnnotationOperationActive = ACTIVE and operationStatus = STARTED state.

Be aware that usually in a single operation run there could be 2K to 5k annotations which could be created. Shoppers can name this API from many alternative machines or threads for quick upserts.



As soon as the annotations have been created in an operation purchasers name FinishAnnotationOperation which modifications following

  • Marks the present operation (let’s say with ID2) to be operationStatus = FINISHED and isAnnotationOperationActive=ACTIVE.
  • We take away the ID2 from the Memcache since it isn’t in STARTED state.
  • Any earlier operation (let’s say with ID1) which was ACTIVE is now marked isAnnotationOperationActive=FALSE in Cassandra.
  • Lastly, we name updateByQuery API in ElasticSearch. This API finds all Elasticsearch paperwork with ID1 and marks isAnnotationOperationActive=FALSE.

Search API

That is the important thing half for our readers. When a shopper calls our search API we should exclude

  • any annotations that are from isAnnotationOperationActive=FALSE operations or
  • for which Annotation operations are at present in STARTED state. We try this by excluding the next from all queries in our system.

To realize above

  1. We add a filter in our ES question to exclude isAnnotationOperationStatus is FALSE.
  2. We question EVCache to search out out all operations that are in STARTED state. Then we exclude all these annotations with annotationId present in memcache. Utilizing memcache permits us to maintain latencies for our search low (most of our queries are lower than 100ms).

Cassandra is our supply of fact so if an error occurs we fail the shopper name. Nonetheless, as soon as we decide to Cassandra we should deal with Elasticsearch errors. In our expertise, all errors have occurred when the Elasticsearch database is having some problem. Within the above case, we created a retry logic for updateByQuery calls to ElasticSearch. If the decision fails we push a message to SQS so we are able to retry in an automatic style after some interval.

In close to time period, we wish to write a excessive stage abstraction single API which could be known as by our purchasers as an alternative of calling three APIs. For instance, they’ll retailer the annotations in a blob storage like S3 and provides us a hyperlink to the file as a part of the one API.