Redis and its function in System Design



Redis is an open-source, in-memory knowledge construction retailer used as a database, cache, and message dealer. It’s extensively used for its quick efficiency, flexibility, and ease of use. 

  • What’s Redis
  • Redis Knowledge Varieties
  • Advantages of utilizing Redis
  • Working Structure of Redis
    • 1. Single Redis Occasion
    • 2. Redis HL (Excessive Availability)
    • 3. Redis Sentinel
    • 4. Redis Cluster / Redis Cluster Grasp-Slave Mannequin
      • What’s gossiping within the Redis cluster?
  • Varieties of Redis Persistence Fashions
    • 1. RDB (Actual-time Knowledge Base) Persistence Mannequin:
      • Snapshotting in RDB
      • Benefits of RDB(Actual-time database)
      • Disadvantages of RDB(Actual-time database)
    • 2. AOF (Append-Solely File) Persistence Mannequin
      • How AOF works?
      • Benefits of AOF
      • Disadvantages of AOF
    • 3. No Persistence Mannequin
    • 4. Hybrid (RDB+AOF) Persistence Mannequin
  • Availability, Consistency, and Partitioning in Redis
  • Can we use Redis as an alternative choice to the unique DB?
  • Conclusion

Redis Knowledge Storage Varieties

Redis permits builders to retailer, retrieve, and manipulate knowledge in varied knowledge buildings comparable to strings, bitmaps, bitfields, hashes, lists, units, sorted units, geospatial, hyperlogs, and streams.

Redis knowledge varieties

Advantages of utilizing Redis

All Redis knowledge resides within the server’s fundamental reminiscence, in distinction to databases comparable to PostgreSQL, SQL Server, and others that retailer most knowledge on disk. Redis can due to this fact assist larger orders of magnitude of operations and quicker response instances. Thus, it leads to super-fast efficiency with common learn and writes operations taking lower than milliseconds, and thus accordingly helps hundreds of thousands of operations per second.

Redis offers an total system that gives us caching programs in each varieties of architectures – monolithic and distributed, thereby making the retrieval of information quicker, because the direct entry operation, by key in reminiscence(like hashtables), will cut back the general complexity of studying the info from the unique SQL Database.

How redis is beneficial for caching

Working Structure of Redis

There are a number of Redis architectures, relying on the use case and scale:

1. Single Redis Occasion

That is probably the most easy Redis deployment. It entails customers establishing and operating small situations that may assist them develop and velocity up their providers. Nonetheless, it has its personal disadvantage, as all calls made to Redis would fail if this operating occasion crashes or is unavailable. Thus there’s a degradation within the total efficiency and velocity of the system.

Single Redis Occasion

2. Redis HA (Excessive Availability)

  • One other widespread setup is the principle deployment with a secondary deployment that’s at all times stored in sync with the replication. The secondary situations might be a number of situations in our deployment, which helps in scale reads from Redis, and supply failover within the case when the principle is misplaced.

Redis HA (secondary failover)

3. Redis Sentinel

  • Sentinel corresponds to a distributed system. It’s designed in a approach the place there’s a cluster of sentinel processes working collectively for coordination of state to offer fixed availability of the Redis system. Listed here are the tasks of the sentinel:
    • Monitoring: Guaranteeing fundamental and secondary situations are working as anticipated.
    • Notification: Notify all of the system admins in regards to the occasions occurring in Redis situations.
    • Administration throughout failure: Sentinel nodes can begin a course of throughout failure if the first occasion isn’t obtainable for lengthy sufficient, and sufficient nodes agree that it’s true.

Redis sentinel

4. Redis Cluster / Redis Cluster Grasp-Slave Mannequin: The Final Structure of Redis

The Redis cluster is the final word structure of Redis. It permits for horizontal scaling of Redis.

In Redis cluster, we determine to unfold the info we’re storing throughout a number of machines, which is named Sharding. So every such Redis occasion within the cluster is taken into account a shard of the entire knowledge.

The Redis Cluster makes use of algorithmic sharding. To seek out the shard for a given key, we hash the important thing and mod the entire end result by the variety of shards. Then, utilizing a deterministic hash operate, which means {that a} given key will at all times map to the identical shard, we will purpose about the place a selected key might be after we learn it sooner or later.

Redis Cluster Structure in System Design

To deal with additional addition of shards into the system (resharding), the Redis cluster makes use of Hashslot, to which the entire knowledge is mapped. Thus, after we add new shards, we merely transfer hashslots from shard to shard and simplify the method of including new main situations into the cluster. And to the benefit, that is attainable with none downtime, and minimal efficiency hit. Let’s take a look at an instance beneath:

Take into account the variety of hashslots to be 10K.
Instance1 comprises hashslots from 0 to 5000
Instance2 comprises hashslots from 5001 to 10000.

Now, let’s say we have to add one other occasion, now the distribution of hashslots involves,

Instance1 comprises hashslots from 0 to 3333.
Instance2 comprises hashslots from 3334 to 6667.
Instance3 comprises hashslots from 6668 to 10000.

What’s gossiping within the Redis cluster?

To find out all the cluster’s well being, the redis cluster makes use of gossiping. Within the instance beneath, we now have 3 fundamental situations and three secondary nodes of them. All these nodes continuously decide which nodes are at the moment obtainable to serve requests. Suppose, if sufficient shards agree that instance1 isn’t responsive, they will promote instance1’s secondary right into a main to maintain the cluster wholesome. As a common rule of thumb, it’s important to have an odd variety of main nodes and two replicas every for probably the most sturdy and fault-tolerant community.

Varieties of Redis Persistence Fashions

Redis offers two fundamental persistence choices to save lots of knowledge to disk: RDB and AOF. Each choices have their very own benefits and drawbacks, and which one to make use of is determined by the precise wants of the applying. Give beneath are the a number of persistence choices listed:

1. RDB (Actual-time Knowledge Base) Persistence Mannequin:

RDB is a point-in-time snapshot of the Redis dataset saved as a binary file. The RDB file comprises a illustration of the dataset at a selected time limit and can be utilized to revive the dataset in case of a server crash or restart. RDB may be very environment friendly when it comes to disk house utilization and efficiency, because it makes use of a binary format to retailer the info.

RDB might be configured to save lots of the info periodically or based mostly on sure circumstances, such at least variety of write operations. Nonetheless, the draw back of RDB is that it could actually result in knowledge loss if the server crashes earlier than the scheduled RDB snapshot is taken.

Snapshotting in RDB

Snapshotting is a course of in Redis persistence that creates a point-in-time snapshot of all the dataset in reminiscence and saves it to disk in a binary format. This snapshot can be utilized to revive the dataset in case of a server crash or restart. Redis helps snapshotting by way of its RDB persistence mechanism.

The snapshotting course of works as follows:

  • Redis forks a baby course of from the father or mother course of.
  • The kid course of creates a duplicate of the present state of the dataset in reminiscence.
  • The kid course of writes the copy of the dataset to a short lived RDB file.
  • The kid course of renames the non permanent file to the ultimate RDB file title, overwriting any present RDB file.
  • The kid course of terminates, and Redis continues serving requests.

Redis might be configured to carry out snapshotting routinely at common intervals or based mostly on sure circumstances, such at least variety of write operations or a minimal period of time elapsed because the final snapshot. If we’re doing heavy work and altering a lot of keys, then a snapshot per minute might be generated for us, in case adjustments are comparatively much less, then a snapshot per 5 minutes, and if it’s additional much less, then each quarter-hour a snapshot might be taken.


Benefits of RDB(Actual-time database)

  • RDB recordsdata are good for backups, as it’s a very compact single-file point-in-time illustration of the redis knowledge. It permits us to simply restore completely different variations of the info set in case of disasters.
  • It is extremely good for catastrophe restoration, being a single compact file that may be transferred to far knowledge facilities.

Disadvantages of RDB(Actual-time database)

Allow us to now take a comparative take a look at the Disadvantages of Redis DB:

  • RDB is not optimum
  • If we have to reduce the prospect of information loss in case Redis stops working. 
  • We will configure completely different save factors the place an RDB is produced. Nonetheless, we’ll often create an RDB snapshot each 5 minutes or extra, so in case of Redis stops working and not using a right shutdown for any purpose, we ought to be ready to lose the most recent minutes of information.

2. AOF (Append-Solely File) Persistence Mannequin

AOF logs all write operations to a file in a human-readable format. This file comprises a document of all of the write operations carried out on the dataset because the final save, making it attainable to reconstruct the dataset in case of a crash. AOF offers higher sturdiness than RDB, because it logs each write operation to disk.

AOF might be configured to save lots of the info periodically or based mostly on sure circumstances, such at least variety of write operations. Nonetheless, AOF can result in slower efficiency and bigger disk house utilization, because it logs each write operation to disk.

The append-only file is another, fully-durable technique for Redis, because the snapshotting isn’t very sturdy.

The AOF might be turned within the configuration file by,

appendonly sure


How AOF works?

  • Redis forks a baby course of from the father or mother course of.
  • The kid course of creates a duplicate of the present state of the dataset in reminiscence.
  • The kid course of writes the copy of the dataset to a brand new AOF in a short lived file.
  • The father or mother course of accumulates all the brand new adjustments in an in-memory buffer (however on the identical time it writes the brand new adjustments within the previous append-only file, so if the rewriting fails, we’re protected).
  • When the kid is completed rewriting the file, the father or mother course of will get a sign and appends the in-memory buffer on the finish of the file generated by the kid.
  • Then, Redis routinely renames the previous file into the brand new one and begins appending new knowledge into the brand new file.

Benefits of AOF

  • AOF Redis is rather more sturdy, as we will have completely different fsync insurance policies, no fsync in any respect, fsync each second, fsync at each question.
  • It’s an append-only log, so there aren’t any seeks, nor corruption issues if there’s energy outage. 
  • The Redis check-of instrument is routinely in a position to repair any half-written command if the log ends abruptly as a result of some disk-full or different causes.

Disadvantages of AOF

Allow us to now take a comparative take a look at the Disadvantages of AOF:

  • These recordsdata are usually larger than equal RDB recordsdata for the identical dataset. 
  • It may be slower than RDB relying on the precise fsync coverage.  
  • AOF can enhance the info consistency however doesn’t assure so probably you’ll be able to lose your knowledge however lower than RDB mode contemplating the RDB is quicker. 

Which one to decide on – Actual-time database (RDB) or Append Solely Recordsdata (AOF)?

The final thought course of ought to be that we use each the persistence strategies if we would like a level of information security akin to what regular databases like PostgreSQL, can present us. If we care loads about our knowledge however nonetheless can dwell with a couple of minutes of information loss in case of disasters, we will merely use RDB alone.

3. No Persistence Mannequin

Redis additionally offers an choice to disable persistence altogether, through which case the info is saved solely in reminiscence. This selection is beneficial when Redis is used as a cache, and the info might be regenerated if misplaced.

4. Hybrid (RDB+AOF) Persistence Mannequin

Redis offers an choice to make use of each RDB and AOF persistence collectively, which is named hybrid persistence. This selection offers the advantages of each RDB and AOF, because the AOF log is used to replay write operations after a restart, and the RDB snapshot is used to revive the dataset at a selected time limit.

Availability, Consistency, and Partitioning in Redis

Right here’s a short overview of how Redis handles availability, consistency, and partitioning:

  • Availability: Redis makes use of a master-slave replication mannequin to make sure excessive availability. This implies that there’s a single “grasp” node that accepts all writes and a number of “slave” nodes that replicate knowledge from the grasp in real-time. Within the occasion of a failure of the grasp node, one of many slave nodes might be promoted to develop into the brand new grasp.
  • Consistency: Redis offers robust consistency ensures for single-key operations, which means that if a worth is written to a key, will probably be instantly obtainable for reads from any node within the cluster. Nonetheless, Redis doesn’t present transactional consistency for multi-key operations, which means that it’s attainable for some nodes to see a special view of the info than others.
  • Partitioning: Redis helps sharding, which permits the info set to be partitioned throughout a number of nodes. Redis makes use of a hash-based partitioning scheme, the place every secret is assigned to a selected node based mostly on its hash worth. Redis additionally offers a mechanism for redistributing knowledge when nodes are added or faraway from the cluster.

Can we use Redis as an alternative choice to the unique DB?

Primarily based on the above dialogue, Redis appears to be a greater choice for the unique DB because it offers quicker retrievals. Even then Redis is not used as a main choice for the database within the system. 

Redis ought to at all times come because the second assist to enhance the efficiency of the general system, as a result of in line with the CAP theorem, Redis is neither constant nor extremely obtainable.

It is because, within the case of the server crashes, we might lose the entire knowledge which is within the reminiscence. It’s okay to lose this knowledge in case of a crash, however for another apps, it turns into actually vital to reload Redis knowledge instantly after the server restarts.


Total, Redis is a robust instrument for system design, but it surely might not be appropriate for all use instances. It is very important rigorously take into account its limitations when deciding whether or not to make use of Redis in a selected software.