In the database fast lane with the Distributed Cache Service

In this article you will read

• what is behind the Distributed Cache Service (DCS),
• where its advantages come into play,
• and what possibilities there are for interactions with a database.

When things need to be done quickly or many users need to be supported at the same time, databases can quickly reach their limits – or the user experience drops drastically. The Distributed Cache Service (DCS) provides a quick remedy. Our whitepaper shows how the DCS can be utilized and uses measurement data to show what users of T Cloud Public (both service providers and end users) can expect from the performance of a DCS.

In most cases, one of the classic relational databases from T Cloud Public portfolio, e.g., a PostgreSQL or MySQL database, will do. So why a cache service? Is it a replacement for a database? The answer – at least to the latter question – is a clear no. The DCS is not a replacement for a classic database or document database. And users who use PostgreSQL to store business-critical data can relax – they don't have to change anything. Whenever data sizes are predictable and limited, data needs to be stored long-term or complex scenarios need to be realized, relational databases are the undisputed queens of the scene.

If it's a matter of milliseconds or many simultaneous users need to be served with a consistently high user experience (i.e. without long waiting times), service providers can consider a DCS. The DCS is (as the name suggests) a distributed, fast online in-memory cache service.  

This means that it stores data in the main memory, not on storage systems like classic relational databases. It is reliable, scalable, ready to use “out of the box” and easy to manage. What sets DCS apart: It offers high read/write performance and fast data access. DCS is compatible with Redis (Remote Dictionary Server), a popular open-source NoSQL data store. It can process a wide variety of data structures such as strings, i.e., texts, binary number sequences, lists, sets, hashes, geodata indices, etc. A complete list of usable data structures can be found in the whitepaper.

Typical application scenarios in which such requirements can arise are, for example, session storing, real-time analysis or machine learning. Session storing is used for many frequently used internet services. Here, the cache service manages the user data required for a visit (a “session”). It provides the special user profile, the credentials and the user-specific personalization. The service thus recognizes the user – in no time at all. 

Another frequent requirement of many internet services is the real-time analysis of data from streams. The problem: streams are characterized by the fact that they do not “stand still”. New data for the stream is continuously generated at close intervals, e.g., from social media, IoT scenarios, or matchmaking. When users visit certain websites, existing data must be “digested” and analyzed quickly so that the visitor receives personalized advertising in a timely manner, for example, or the service makes him/her a suitable offer based on a real-time decision.

In our whitepaper, we used three use cases to measure how increasing user numbers affect the performance of an online service – with and without DCS: the availability of frequently requested data, the storage of “provisional” data when creating a shopping cart, and the processing of temporary data to be analyzed for a comment. The data clearly shows how response times remain stable with the DCS, whereas they decrease when using a relational database. Overall, the DCS offers 5 to 30 times the speed.

However, because data also needs to be stored permanently, service providers cannot avoid a cost-effective standard database. This often results in hybrid database scenarios in which the relational database is the “packhorse” for continuous operation and the DCS acts as a reactive companion (perhaps a kind of greyhound) that quickly tackles or pre-processes anything out of the ordinary. Our whitepaper shows five principle architectures in which the DCS can interact with a classic database.

The DCS in T Cloud Public is currently based on Redis. Redis 7.x will be rolled out as an update in the first quarter of 2025. Redis 3, on the other hand, will be retired in mid-2025. Users of the old version should therefore consider migrating to a newer version.