Home Big Data Apache Ozone – A Multi-Protocol Conscious Storage System

Apache Ozone – A Multi-Protocol Conscious Storage System

Apache Ozone – A Multi-Protocol Conscious Storage System


Are you struggling to handle the ever-increasing quantity and number of information in at present’s continually evolving panorama of contemporary information architectures? The huge tapestry of information sorts spanning structured, semi-structured, and unstructured information means information professionals must be proficient with numerous information codecs reminiscent of ORC, Parquet, Avro, CSV, and Apache Iceberg tables, to cowl the ever rising spectrum of datasets – be they photographs, movies, sensor information, or different sort of media content material. Navigating this intricate maze of information will be difficult, and that’s why Apache Ozone has turn into a preferred, cloud-native storage resolution that spans any information use case with the efficiency wanted for at present’s information architectures.  

Apache Ozone, a extremely scalable, excessive efficiency distributed object retailer,  offers the perfect resolution to this requirement with its bucket structure flexibility and multi-protocol assist. Apache Ozone is appropriate with Amazon S3 and Hadoop FileSystem protocols and offers bucket layouts which can be optimized for each Object Retailer and File system semantics. With these options, Apache Ozone can be utilized as a pure object retailer, a Hadoop Appropriate FileSystem (HCFS), or each, enabling customers to retailer several types of information in a single retailer and entry the identical information utilizing a number of protocols offering the scale of an object retailer and the flexibleness of the Hadoop File system.   

A earlier weblog put up describes the totally different bucket layouts accessible in Ozone. This weblog put up is meant to supply steering to Ozone directors and utility builders on the optimum utilization of the bucket layouts for various functions.

To start out with, Ozone’s namespace contains the next conceptual entities:

Fig.1 Apache Ozone Namespace structure

  • Volumes are the highest degree namespace grouping in Ozone. Quantity names should be distinctive and can be utilized for tenants or customers. 
  • Buckets can be utilized as guardian directories. Every quantity can include a number of buckets of information. Bucket names should be distinctive inside a quantity. 
  • Keys retailer information inside buckets. Keys will be information, directories, or objects.

Bucket Layouts in Apache Ozone

File System Optimized (FSO) and Object Retailer (OBS) are the 2 new bucket layouts in Ozone for unified and optimized storage in addition to entry to information, directories, and objects. Bucket layouts present a single Ozone cluster with the capabilities of each a Hadoop Appropriate File System (HCFS) and Object Retailer (like Amazon S3). Considered one of these two layouts must be used for all new storage wants.

An outline of the bucket layouts and their options are under.

Fig 2. Bucket Layouts in Apache Ozone

Interoperability between FS and S3 API

Customers can retailer their information in Apache Ozone and may entry the information with a number of protocols. 

Protocols supplied by Ozone:

  • ofs
    • ofs is a Hadoop Appropriate File System (HCFS) protocol.
  • ozone fs is a command line interface just like “hdfs dfs” CLI that works with HCFS protocols like ofs.
    • Most conventional analytics functions like Hive, Spark, Impala, YARN and many others. are constructed to make use of the HCFS protocol natively and therefore they will use the ofs protocol to entry Ozone out of the field with no modifications.
    • Trash implementation is obtainable with the ofs protocol to make sure secure deletion of objects.
  • S3
    • Any cloud-native S3 workload constructed to entry S3 storage utilizing both the AWS CLI, Boto S3 consumer, or different S3 consumer library can entry Ozone by way of the S3 protocol.
    • Since Ozone helps the S3 API, it can be accessed utilizing the s3a connector. S3a is a translator from the Hadoop Appropriate Filesystem API to the Amazon S3 REST API.
    • Hive, Spark, Impala, YARN, BI instruments with S3 connectors can work together with Ozone utilizing the s3a protocol.
    • When accessing FSO buckets by way of the S3 interface, paths are normalized, however renames and deletes are not atomic.
    • s3a will translate listing renames to particular person object renames on the consumer earlier than sending them to Ozone. Ozone’s S3 gateway will ahead the article renames to the FSO bucket.
    • Entry to LEGACY buckets utilizing S3 interface is identical as entry to FSO bucket if, ozone.om.allow.filesystem.paths=true in any other case, it’s the identical as entry to OBS bucket.
  • o3
    • Ozone Shell (ozone sh) is a command line interface used to work together with Ozone utilizing the o3 protocol.
    • Ozone Shell is really helpful to make use of for quantity and bucket administration, however it can be used to learn and write information.
    • Solely anticipated for use by cluster directors.

Fig 3. Interoperability between FS and S3 APIOzone’s assist for interoperability between File System and Object Retailer API can facilitate the implementation of hybrid cloud use circumstances reminiscent of:

1- Ingesting information utilizing S3 interface into FSO buckets for low latency analytics utilizing the ofs protocol. 

Fig 4. Ingest utilizing S3 API and devour utilizing FS API

2- Storing information on-premises for safety and compliance which can be accessed utilizing cloud-compatible API.

Fig 5. Ingest utilizing FS API and devour utilizing S3 API

When to make use of FSO vs OBS Bucket Layouts

Fig 6. When to make use of FSO vs OBS

  • Analytics companies constructed for HDFS are significantly nicely fitted to FSO buckets:
  • Apache Hive and Impala drop desk question, recursive listing deletion, and listing shifting operations on information in FSO buckets are quicker and constant with none partial leads to case of any failure as a result of renames and deletes are atomic and quick.
  • Job Committers of Hive, Impala, and Spark usually rename their momentary output information to a closing output location on the finish of the job. Renames are quicker for information and directories in FSO buckets.
  • Cloud-native functions constructed for S3 are higher fitted to OBS buckets:
  • OBS buckets present strict S3 compatibility.
  • OBS buckets present wealthy storage for media information and different unstructured information enabling exploration of unstructured information.


Bucket layouts are a strong function that enable Apache Ozone for use as each an Object Retailer and Hadoop Appropriate File System. On this article, we’ve got lined the advantages of every bucket structure and the way to decide on the very best bucket structure for every workload.

In case you are focused on studying extra about methods to use Apache Ozone to energy information science, this is a superb article. If you wish to know extra about Cloudera on non-public cloud, see right here.

Our Skilled Providers, Help and Engineering groups can be found to share their data and experience with you to decide on the correct bucket layouts in your numerous information and workload wants and optimize your information structure. Please attain out to your Cloudera account crew or get in contact with us right here.


[1] https://weblog.cloudera.com/apache-ozone-a-high-performance-object-store-for-cdp-private-cloud/

[2] https://weblog.cloudera.com/a-flexible-and-efficient-storage-system-for-diverse-workloads/



Please enter your comment!
Please enter your name here