Introducing Dragonfly Cloud! Learn More

Question: How fast is an in-memory database?


In-memory databases (IMDBs) are significantly faster than disk-based databases due to their fundamental design principle of storing data in main memory (RAM) as opposed to on traditional disk storage. The speed of IMDBs can range from tens of thousands to millions of transactions per second, depending on the specific workload and configuration.

Factors contributing to this speed include:

  1. No Disk I/O: Unlike disk-based databases, IMDBs do not need to read/write data from/to disk. Disk I/O operations are generally slow and are often a bottleneck in data processing. By eliminating these operations, IMDBs drastically reduce latency and increase throughput.

  2. Optimized Data Structures: IMDBs often use data structures specifically optimized for memory access. Traditional B-trees or other disk-optimized data structures may not be efficient in memory; thus, IMDBs might use variations like T-trees, hash maps, or skip lists.

  3. Concurrency Control: Many IMDBs offer multi-version concurrency control (MVCC), allowing multiple transactions to occur concurrently without locking resources, which can improve speed and efficiency.

  4. Simplified Software Stack: The software stack in an IMDB is typically simpler than that of a disk-based database, resulting in fewer CPU instructions per operation, which improves performance.

Here's an example using Redis, a popular in-memory data store:

import redis r = redis.Redis() # Writing 10000 key-value pairs for i in range(10000): r.set(f'key{i}', f'value{i}') # Reading 10000 key-value pairs for i in range(10000): value = r.get(f'key{i}')

In the code snippet above, you'll notice that reading/writing 10,000 key-value pairs occurs almost instantaneously, showcasing the speed of in-memory databases.

However, it's important to note that while IMDBs are incredibly fast, they also come with certain trade-offs such as cost (RAM is more expensive than disk storage) and volatility (data in RAM is lost if the system crashes or is shut down, unless strategies like persistence or replication are employed). Hence, the decision to use an in-memory database should be based on your specific use case.

Was this content helpful?

White Paper

Free System Design on AWS E-Book

Download this early release of O'Reilly's latest cloud infrastructure e-book: System Design on AWS.

Free System Design on AWS E-Book

Start building today 

Dragonfly is fully compatible with the Redis ecosystem and requires no code changes to implement.