Architecture Patterns: Caching(Part 1)

The first steps to understanding caching

Kislay Verma

--

Originally published on my website — https://kislayverma.com/software-architecture/architecture-patterns-caching-part-1/

Performance has always been a key feature of technical systems. Today on the internet, sub-second latencies are the norm. It costs companies money if their pages load slowly because potential customers won’t wait longer than that. On the other hand, there is more and more data from many different sources which have to be loaded into a rich user experience (think of the number of things going on a typical FB page). This data gathering problem is further exacerbated by the trend towards microservices.

Given all this, how are we to build super-fast systems?

What is caching?

Caching is the general term used for storing some frequently read data temporarily in a place from where it can be read much faster than reading it from the source (database, file system, service whatever). This reduction in data reading time reduces the system’s latency. This also increases the system’s overall throughput because requests are being served faster and hence more requests can be served per unit time.

Microprocessor architectures have long employed this technique to make programs run faster. Instead of reading all data from the file system all the time, microprocessors employ multiple levels of cache where it is slower to read from lower levels than from the higher levels. The game is now one of making sure that the data read most often is in the highest level cache, and so on. Getting this right can make a dramatic difference to the speed of a running program.

Typically, caches hold a copy of the source data for some time (called expiration time or time-to-live (TTL)) after which the data is “evicted” from the cache. As more data is loaded into a cache of finite capacity, some strategies may be applied to decide which data is retained and which is evicted.

Where can we employ caching

  1. The system must be read-heavy — As should be clear by now that caching is only a solution for scaling the reading of data. So if you are building a system that reads data…

--

--