“Cache Management in Microservices” is locked Cache Management in Microservices

0 7

By ravurisuman Microservices June 8, 2018

Caching is one of the standard and proven technique to improve the turnaround time for a given service or application. In traditional application development, we have different patterns to follow to build, manage and serve response from cache. In the context of Microservices, some of those patterns are not so relevant and demands us to be more creative. This blog talks about different caching types and their pros & cons and recommended caching type for Microservices environment.

Why caching

It is anonymous across design community that at least few of the services in any enterprise application needs caching enabled, to ensure a better turnaround. Caching is a technique where data gets returned directly from service, rather than making a call to another application (can be a database or another service) or recomputing the value which has already computed once. We can reap very good performance benefits, if we apply right caching techniques for the right data in a given context.

Caching types

Based on the approach we follow to warm the cache, we have two types of caches:
- Preloaded cache: Load the data in cache before the start of service and serve data from cache from the first request.
- Lazy loaded cache: Warm the cache as data is being requested i.e. first request for a specific data will hit the server or compute, then from next request onwards, data will be served from cache.

Based on where do you maintain the cached data, we have two types of caches:
- Inservice cache: Cache stays with in the memory of service instance.
- Distributed cache: Cache is stored outside the service instance but accessible to all the service instances of given service.

Challenges with Microservices environment

Microservices are inherently designed and expected to scale with ease.

Cache Type	Inservice Cache	Distributed Cache
Preloaded Cache	Pros: Consumer app will start seeing the performance gain from the time service is launched. New service launched as part of autoscale will start having preloaded cache, hence it’s good for microservices kind of environment Cons: Launching of new service may take longer as cache needs to be preloaded. Service may become fat as we might be caching all the possible data irrespective of frequency of usage	Pros: Consumer app will start seeing the performance gain from the time service is launched. New service launched as part of autoscale will start having preloaded cache, hence it’s good for microservices kind of environment Cons: Populating distributed cache upfront may be a challenge. We may need to build and/or use separate tool to preload the data
Lazy Loaded Cache	Pros: Launching of new service will take less time as there is no burden of preloading Cons: Initial set of users will have slower performance Scale without losing on performance is tricky as newly launched instance may take time to warm up the cache	Pros: Launching of new service will take less time as there is no burden of preloading New service launched as part of autoscale will start having access to distributed cache, hence good for microservices kind of environment Cons: Due to any reason, if distributed cache fails overall service performance gets degrade in a big way

Cache Type

Inservice Cache

Distributed Cache

Preloaded Cache

Pros:

Consumer app will start seeing the performance gain from the time service is launched.
New service launched as part of autoscale will start having preloaded cache, hence it’s good for microservices kind of environment

Cons:

Launching of new service may take longer as cache needs to be preloaded.
Service may become fat as we might be caching all the possible data irrespective of frequency of usage

Pros:

Consumer app will start seeing the performance gain from the time service is launched.
New service launched as part of autoscale will start having preloaded cache, hence it’s good for microservices kind of environment

Cons:

Populating distributed cache upfront may be a challenge. We may need to build and/or use separate tool to preload the data

Lazy Loaded Cache

Pros:

Launching of new service will take less time as there is no burden of preloading

Cons:

Initial set of users will have slower performance
Scale without losing on performance is tricky as newly launched instance may take time to warm up the cache

Pros:

Launching of new service will take less time as there is no burden of preloading
New service launched as part of autoscale will start having access to distributed cache, hence good for microservices kind of environment

Cons:

Due to any reason, if distributed cache fails overall service performance gets degrade in a big way

Cache Triangle

Summary

As part of this article, I tried to cover different aspects of caching, pros & cons of each approach. I would recommend distributed and lazy loaded caching for Microservices environment, as it will help us to bring the right balance among three core aspects of caching. With the advent of public cloud, managing distributed cache is relatively easy. If we choose lazy loading, performance for the initial requests may not be great and if that becomes serious concern then we can handle that with instances of high I/O throughput and/or computing power or by optimizing certain part of functional logic. With this article I hope you gained certain knowledge which will help you to take an informed decision while choosing caching technique in your Microservices environment.