Apache HBase VS Scylla

### YouTube The popular implementations of an on-demand video streaming service are the following: - YouTube - Netflix - Vimeo - TikTok --- #### Requirements - The user (**client**) can upload video files - The user can stream video content - The user can search for videos based on the video title --- #### Data storage ##### Database schema - The primary entities are the videos, the users, and the comments tables - The relationship between the users and the videos is 1-to-many - The relationship between the users and the comments table is 1-to-many - The relationship between the videos and the comments table is 1-to-many --- ##### Type of data store - The wide-column data store ([LSM](https://en.wikipedia.org/wiki/Log-structured\_merge-tree) tree-based) such as [Apache HBase](https://hbase.apache.org/) is used to persist thumbnail images for clumping the files together, fault-tolerance, and replication - A cache server such as Redis is used to store the metadata of popular video content - Message queue such as Apache Kafka is used for the asynchronous processing (encoding) of videos - A relational database such as MySQL stores the metadata of the users and the videos - The video files are stored in a managed object storage such as AWS S3 - Lucene-based inverted-index data store such as Apache Solr is used to persist the video index data to provide search functionality --- #### High-level design - Popular video content is streamed from CDN - Video encoding (**transcoding**) is the process of converting a video format to other formats (MPEG, HLS) to provide the best stream possible on multiple devices and bandwidth - A message queue can be configured between services for parallelism and improved fault tolerance Codecs (H.264, VP9, HEVC) are compression and decompression algorithms used to reduce video file size while preserving video quality - The popular video streaming protocols (data transfer standard) are **MPEG-DASH** (Moving Pictures Experts Group - Dynamic Adaptive Streaming over HTTP), **Apple HLS** (HTTP Live Streaming), **Microsoft Smooth Streaming**, and **Adobe HDS** (HTTP Dynamic Streaming) --- #### Video upload workflow 1. The user (**client**) executes a DNS query to identify the server 2. The client makes an HTTP connection to the load balancer 3. The video upload requests are rate limited to prevent malicious clients 4. The load balancer delegates the client's request to an API server (**web server**) with free capacity 5. The web server delegates the client's request to an app server that handles the API endpoint 6. The ID of the uploaded video is stored on the message queue for asynchronous processing of the video file 7. The title and description (**metadata**) of the video are stored in the metadata database 8. The app server queries the object store service to generate a pre-signed URL for storing the raw video file 9. The client uploads the raw video file directly to the object store using the pre-signed URL to save the system network bandwidth 10. The transcoding servers query the message queue using the publish-subscribe pattern to get notified on uploaded videos 11. The transcoding server fetches the raw video file by querying the raw object store 12. The transcoding server transcodes the raw video file into multiple codecs and stores the transcoded content on the transcoded object store 13. The thumbnail server generates on average five thumbnail images for each video file and stores the generated images on the thumbnail store 14. The transcoding server persists the ID of the transcoded video on the message queue for further processing 15. The upload handler service queries the message queue through the publish-subscribe pattern to get notified on transcoded video files 16. The upload handler service updates the metadata database with metadata of transcoded video files 17. The upload handler service queries the notification service to notify the client of the video processing status 18. The database can be partitioned through [consistent hashing](https://systemdesign.one/consistent-hashing-explained/) (key = user ID or video ID) 19. [Block matching](https://en.wikipedia.org/wiki/Block-matching\_algorithm) or [Phase correlation](https://en.wikipedia.org/wiki/Phase\_correlation) algorithms can be used to detect the duplicate video content 20. The web server (API server) must be kept stateless for scaling out through replication 21. The video file is stored in multiple resolutions and formats in order to support multiple devices and bandwidth 22. The video can be split into smaller chunks by the client before upload to support the resume of broken uploads 23. Watermarking and encryption can be used to protect video content 24. The data centers are added to improve latency and data recovery at the expense of increased maintenance workflows 25. Dead letter queue can be used to improve fault tolerance and error handling 26. Chaos engineering is used to identify the failures on networks, servers, and applications 27. Load testing and chaos engineering are used to improve fault tolerance 28. [RAID](https://en.wikipedia.org/wiki/RAID) configuration improves the hardware throughput 29. The data store is partitioned to spread the writes and reads at the expense of difficult joins, transactions, and fat client 30. Federation and sharding are used to scale out the database 31. The write requests are redirected to the leader and the read requests are redirected to the followers of the database 32. [Vitess](https://vitess.io/) is a storage middleware for scaling out MySQL 33. Vitess redirects the read requests that require fresh data to the leader (For example, update user profile operation) 34. Vitess uses a lock server (Apache Zookeeper) for automatic sharding and leader election on the database layer 35. Vitess supports RPC-based joins, indexing, and transactions on SQL database 36. Vitess allows to offload of partitioning logic from the application and improves database queries by caching

HBase, A scalable, distributed database that supports structured data storage for large tables.

Today, it is normal for enterprises to leverage diversified databases. In my market of expertise, China, in the Internet industry, MySQL together with data sharding middleware is the go to architecture, with GreenPlum, HBase, Elasticsearch, Clickhouse and other big data ecosystems being auxiliary computing engine for analytical data. At the same time, some legacy systems (such as SQLServer legacy from .NET transformation, or Oracle legacy from outsourcing) can still be found in use. In the financial industry, Oracle or DB2 is still heavily used as the core transaction system. New business is migrating to MySQL or PostgreSQL. In addition to transactional databases, analytical databases are increasingly diversified as well.

For this type of systems PostgreSQL not best solution, and for a number of reasons like lack of replication out of the box. And we strictly must not have «Vendor lock», and therefore also did not take modern SQL databases like Amazon Aurora. And end of the ends the choice was made in favor Cassandra, for this article where we will talking about low-lever implementation of Repository Pattern it is not important, in your case it can be any unpopular database like HBase for example.

Apache HBase

Wide column: Apache Cassandra, Apache HBase

The authors performed an analysis of critical failures of the five distributed systems: Cassandra, HBase, HDFS, MapReduce, and Redis.

Scylla which is a C++ implementation of the Cassandra distributed K:V store https://github.com/scylladb/scylladb

What you say is impossible, we pretty successfully apply at ScyllaDB (see https://github.com/scylladb/scylladb/commits/master).

I'm not sure 100% of the commits compile & pass all tests - there may be some mistakes - but generally we're in a pretty good state, and the clean git log is being successfully used for bisecting.

If you want even larger scale - if I understand correctly, the Linux kernel practices a similar thing, which is where we got this practice from (ScyllaDB founders came from kernel development). And since Git was originally created to help developing Linux - that's where you want to look for good practices.

Examples: - Vitess - MySQL cluster - YugabyteDB - ScyllaDB - Couchbase - ArangoDB

Never used it before but check out https://github.com/scylladb/scylla

I have a lot of experience with Cassandra though.

I posted this because I'm interested to hear from anyone using it - how has it worked out for you?

I note it's written in C++ which is a bit of a surprise - I'd expected Rust or Golang.

Interesting as well is is AGPL - licensing is always contentious:

https://github.com/scylladb/scylla/blob/master/LICENSE.AGPL

Failure detection is now done directly by nodes pinging each other rather than through the gossip protocol. This is more reliable and the information is available more rapidly. Impact on networking is low, since Scylla implements a fully connected mesh in all clusters smaller than 256 nodes per datacenter, which is much larger than the typical cluster.