Disperser

Dispersal

Requesters that want to store data on 0G Storage make requests to the disperser with the form of DisperseBlobRequest.

They specify the data they want to store on 0G Storage. The disperser takes each DisperseBlobRequest and stores it into the s3 bucket. The data that the disperser stores into the s3 bucket contains two parts: blob metadata and full blob data.

Blob Data

The key of the blob data is calculated by a certain hash function. This key is used for users to retrieve the blob directly from a disperser. The disperser uploads the blob data with the object key of the blob hash to a certain s3 bucket (defined by the disperser service).

In pseudocode:

blobHash := getBlobHash(blob)
metadataHash := getMetadataHash(requestedAt, blob.RequestHeader.SecurityParams)
metadataKey.BlobHash = blobHash
metadataKey.MetadataHash = metadataHash

err = s.s3Client.UploadObject(ctx, s.bucketName, blobObjectKey(blobHash), blob.Data)

Metadata

The metadata of a blob is constructed and stored into a table (defined by the disperser service) in aws dynamodb which is a nosql database. The update of the metadata in the dynamodb is monitored by the Batcher service to do further process.

Retrieval

Requesters can directly download the data blob from the disperser service with the form of RetrieveBlobRequest.

The user defines the index of the blob in a batch and the hash of the batch header to retrieve the metadata of the blob in dynamodb first. The disperser then downloads the full blob data using the blob key in the metadata.

Note that the disperser is not responsible for encoding/decoding of the blob. The disperser service is trustless, whether to trust the disperser depends on the user judgement.