This week’s issue brings you:

• Caching Strategies: A Comparative Overview

• Batch Processing vs Real-time Streaming (Recap)

• ELT Clearly Explained (Recap)

• MVC vs MVP — What’s the Difference? (Recap)

READ TIME: 5 MINUTES

Caching Strategies: A Comparative Overview

In the current world of big data and high-speed applications, performance is a prominent consideration for any development team.

Caching is one of the most used techniques to boost performance due to its simplicity and wide range of use cases.

With caching, data is copied and stored in locations that are quick to access such as on the browser or a CDN.

A core part of any caching strategy is how data is updated and evicted from the cache. There are several approaches to choose from, each with distinct use cases.

Least Recently Used (LRU) is an approach to cache management that frees up space for new data by removing data that has not been accessed or utilized for the longest period of time.

It assumes that recently accessed data will be needed again soon.

This is quite a common approach and is often used in browsers, CDNs, and operating systems.

Most Recently Used (MRU) is the opposite of LRU, where the most recently used data is removed first. 

MRU tends not to be used often. However, it can be useful in tightly scoped use cases where recently accessed data is unlikely to be reused—such as certain streaming pipelines or stack-based workflows.

Least Frequently Used (LFU) removes data that is used the least.

Although it is a more accurate approach than LRU, it requires a mechanism to keep count of how often data is accessed, which adds complexity.

LFU also has the risk of keeping outdated data in the cache.

For these reasons, it is often used in combination with other strategies such as LRU.

With Time-To-Live (TTL), data is kept in the cache for a pre-defined period of time.

This is ideal for cases where the current state of data is only valid for a certain period of time, such as session data.

Two-tiered caching provides a more complex approach that strikes a balance between speed and cost. In this design, data is split up between a first and second tier.

This first tier is a smaller, faster, and often more expensive caching tier that stores frequently used data.

The second tier is a larger, slower, and less expensive tier that stores data that is used less frequently.

The five strategies mentioned above are the most popular approaches to caching. There are other notable mentions, such as the following:

• First In, First Out (FIFO): The oldest data is deleted first.

• Random Replacement (RR): Randomly selects data to be deleted.

• Adaptive Replacement Cache (ARC): Uses a self-tuning algorithm that tracks recency and frequency to determine which data to delete first.

There’s no one-size-fits-all solution. The best caching strategy depends on your system’s access patterns, update frequency, latency requirements, and cost constraints. Understanding and combining multiple strategies often leads to the most optimal performance.

Batch Processing vs Real-time Streaming (Recap)

Two approaches for your data pipelines, one goal: turning raw data into real insights.

Batch processing:
Processes large datasets at intervals (eg; daily reports, weekly analytics, ML training).

✅ High throughput
✅ Easy to debug
✅ Cost-effective for scheduled jobs
❌ Not suitable for real-time use cases

Real-time streaming:
Processes data instantly (eg; fraud detection, live dashboards, IoT).

✅ Low latency (milliseconds)
✅ Instant insights and reactions
✅ Scales with high-velocity data
❌ More complex and resource-intensive

Key tradeoffs:
Batch = Simpler, slower, cheaper
Streaming = Faster, more complex, higher ops overhead

Check out the full post for a deeper explanation and breakdown of how each approach works.

ELT Clearly Explained (Recap)

ELT (Extract, Load, Transform) is a modern approach to data processing that reverses the traditional ETL model. Instead of transforming data before storage, ELT first loads raw data into cloud warehouses (like Snowflake or BigQuery), then transforms it on-demand and at scale.

The ELT process explained:

Extract — Gather raw data from diverse sources (eg; APIs, databases, IoT devices). Tools like Fivetran and Airbyte automate this with efficient methods like Change Data Capture (CDC).

Load — Store raw data in a cloud warehouse without altering it. This keeps the data flexible, reusable, and queryable immediately.

Transform — Clean, join, aggregate, and apply business logic using SQL or tools like dbt, all within the warehouse.

For a deeper explanation, see the full post here.

MVC vs MVP — What’s the Difference? (Recap)

• MVC (Model View Controller) and MVP (Model View Presenter) are design patterns that separate an application into distinct components.

• MVP was established after MVC to improve on the drawbacks of MVC and improve maintainability.

• Both patterns aim to separate concerns, but they have some differences in their approach.

• The most significant procedure differences relate to which component handles the business logic and how the UI gets updated.

• MVC is typically less complex but it can have tightly coupled components. Whereas, MVP is generally more complex but has more decoupled components.

Check out the full post for a deeper explanation and comparison of the two design patterns.

That wraps up this week’s issue of Level Up Coding’s newsletter!

Join us again next week where we’ll explore and visually distill more important engineering concepts.