System Design – Design Privacy Setting

Posted on by user

On facebook, you may have noticed that you can set different privacy levels for the posts you published to be only to a specific group of users, like public, friends, friends of friends, etc

Key Features
1. Enable a user to specify the different levels of privacy for a post so that it is only visible to a particular set of users on Facebook
2. For simplicity, implement two levels of privacy , public or Friends, more complexity levels: friends of friends, and custom groups.

Design Goals

  1. Minimum latency and high consistency
  2. Due to CAP, lower the availability in the interest of high consistency
  3. Read-heavy service

Scale Estimation

  1. One billion daily active users(DAUs)
  2. Privacy level will be set by users for about two billion posts daily.
  3. Check the privacy level of about 100 billion posts daily so that they can be made visible to the correct users.

API Design
1. /setPrivacyLevel: api to set the privacy level for a post,
Params: user_id, post_id, privacy_level_enum, and timestamp
2. /canView: api to check if a user can view a particular post or not.
Params: viewer_id, post_id, and timestamp.

High Level Design


To define the different privacy levels, we will user the Enum datatype: 1. public, 2. friends, 3. friends of friends, 4. customized group. This enum will be stored a long each post in the database.

  1. To check if a post is visible to a specific user or not, review the privacy level enum:
    – If it is set to Public, then it will be visible to everyone
    – If it is set to Friends, we will check if the current user is a friend of the post’s owner, if yes, then the post will be displayed on the UI.
  2. Store the friends data of the users in a key-value store.
    – Key will be the user_id, and the value will be a set of all the friends that user has.
    – Since this data will be massive, it will be shared across multiple servers using the user_id as the key
    – Also need to discuss what will happen if a shared dies and create a fault-tolerant design.
  3. Friends of Friends
    – Given the scale of facebook, we can not enumerate through all the friends of friends as it will be extensive list and increase the system’s latency.
    – One possible way to solve this challenging problem is to perform an interaction of the post’s owner and the viewers friends lists.
    – if the intersection results in a non-empty set, then the post can be displayed.
    (from: https://www.youtube.com/watch?v=-LjbhtswNwE&t=335s)

    Scale the K-V Database
    1. Sharding:     Distributing data across multiple machines is crucial for scalability. Sharding can be based on user ID or other unique identifiers, ensuring an even distribution of data and workload.
    2.Cluster master-replica model: To remain available when a subset of master nodes are failing or are not able to communicate with the majority of nodes, let KV Cluster uses a master-replica model where every hash slot has from 1 (the master itself) to N replicas (N-1 additional replica nodes). (source: https://redis.io/docs/management/scaling/)

    Redis Cluster automatically manages sharding and replication. It distributes data across multiple nodes, and each shard’s data is replicated to one or more other nodes.

    In the event of a node failure, Redis Cluster can continue to operate as long as at least one replica of the data is available.

    Redis Cluster also provides some degree of automated failover. If a primary node fails, one of its replicas is automatically promoted to be the new primary.
+-----------------+       +-----------------+       +-----------------+
|     Shard 1     |       |     Shard 2     |       |     Shard 3     |
|   Primary Node  |       |   Primary Node  |       |   Primary Node  |
+-----------------+       +-----------------+       +-----------------+
       /|\                       /|\                       /|\
        |                         |                         |
        | Replication             | Replication             | Replication
        |                         |                         |
       \|/                       \|/                       \|/
+-----------------+       +-----------------+       +-----------------+
| Replica Node 1  |       | Replica Node 1  |       | Replica Node 1  |
+-----------------+       +-----------------+       +-----------------+
+-----------------+       +-----------------+       +-----------------+
| Replica Node 2  |       | Replica Node 2  |       | Replica Node 2  |
+-----------------+       +-----------------+       +-----------------+
9