Redis vs Valkey performance

Summary

In this video, the performance of Redis and Valkey, an open-source fork of Redis, is compared through various tests on virtual machines with single and multi-threaded configurations. The tests measure throughput, latency, CPU usage, memory usage, and network performance. In single-threaded mode, Valkey slightly outperforms Redis, but the most significant difference is observed in multi-threaded tests where Valkey achieves higher throughput and better resource efficiency. The video concludes with a recommendation to switch to Valkey, as it offers better overall performance, is open-source, and has a larger community of contributors.

Takeaways

• 😀 Redis and Valkey were compared in terms of latency, throughput, CPU usage, memory usage, and network performance.
• 😀 In the first test with a single CPU, Redis reached 100,000 requests per second, while Valkey was slightly more stable, reaching 107,000 requests per second.
• 😀 Valkey outperformed Redis in latency, especially when measured with the p90 percentile, in both SET and GET operations.
• 😀 CPU usage was generally similar between Redis and Valkey, with some spikes, but no major differences.
• 😀 Memory and network usage were also measured for both databases, revealing no significant differences.
• 😀 The second test with 4 CPUs highlighted the bigger differences between Redis and Valkey, especially in multithreading performance.
• 😀 Redis reached 260,000 requests per second when using 4 CPUs, while Valkey achieved 350,000 requests per second, showing a notable performance gap.
• 😀 Valkey utilizes resources more efficiently than Redis, especially when both were configured to use 4 I/O threads for operations.
• 😀 In terms of overall performance, Valkey performed better than Redis in all tested areas, including latency and throughput.
• 😀 Despite Redis having new features like JSON support and time-series databases in version 8.0, Valkey is a better overall option for most users due to its higher performance, open-source nature, and larger contributor community.

Q & A

What is the main difference between Redis and Valkey?

The main difference is that Valkey is an open-source fork of Redis, and it implements multithreading independently after the fork, unlike Redis which traditionally uses a single-threaded mode for core operations.

What tests were conducted to compare Redis and Valkey?

Two tests were conducted: one using a virtual machine with a single CPU (single-threaded mode) and another using a virtual machine with 4 CPUs where both databases were configured to use 4 I/O threads.

Which database performed better in the single-threaded test?

In the single-threaded test (single CPU), Valkey outperformed Redis with slightly better throughput, reaching 107,000 requests per second compared to Redis's 100,000 requests per second.

How was latency measured during the tests?

Latency was measured from the client side using the p90 percentile, which is a common way to measure response times for 90% of the requests.

What was the observed difference in CPU usage between Redis and Valkey?

In the single-threaded test, CPU usage was similar for both Redis and Valkey, with some spikes observed. However, in the multi-threaded test (using 4 CPUs), Valkey showed more efficient CPU resource utilization.

How did Redis and Valkey compare in terms of memory usage?

Memory usage was similar for both Redis and Valkey, with no significant differences observed in the tests.

What was the impact of multithreading on performance in the second test?

In the second test with 4 CPUs and 4 I/O threads configured, Valkey significantly outperformed Redis, achieving 350,000 requests per second compared to Redis's 260,000 requests per second.

What configuration was used for the multi-threaded test?

The multi-threaded test used m7a.xlarge instances with 4 CPUs each. Both Redis and Valkey were configured with 4 I/O threads for I/O operations.

Why was the presenter convinced to switch from Redis to Valkey?

The presenter was convinced to switch to Valkey due to its higher throughput, lower latency, and better resource efficiency (especially CPU usage) in the multi-threaded test.

What are the main reasons the presenter gives for switching to Valkey over Redis?

The main reasons include Valkey's superior performance in multi-threaded environments, its open-source nature, and its larger community of contributors. The presenter also notes that unless specific Redis 8.0 features like JSON support or time-series databases are needed, Valkey is the better option.