Read this in other languages: 日本語
This directory contains example code illustrating how to generate load against Momento. It also includes an example that shows how request coalescing can reduce latencies and improve throughput for workloads that it is suitable for.
- Node version 16 or higher is required
- To get started with Momento you will need a Momento API key. You can get one from the Momento Console.
To run any of the examples you will need to install the dependencies once first:
npm installThis project includes a very basic load generator to allow you to experiment with performance in your environment based on different configurations. It's very simplistic, and only intended to give you a quick way to explore the performance of the Momento Cache client running on a single nodejs process.
Note that because nodejs javascript code runs on a single thread, the limiting factor in request throughput will often be CPU. Keep an eye on your CPU consumption while running the load generator, and if you reach 100% of a CPU core then you most likely won't be able to improve throughput further without running additional nodejs processes.
CPU will also impact your client-side latency; as you increase the number of concurrent requests, if they are competing for CPU time then the observed latency will increase.
Also, since performance will be impacted by network latency, you'll get the best results if you run on a cloud VM in the same region as your Momento cache.
Check out the configuration settings at the bottom of the 'load-gen.ts' to see how different configurations impact performance.
If you have questions or need help experimenting further, please reach out to us!
To run the load generator:
# Run example load generator
MOMENTO_API_KEY=<YOUR API KEY> npm run load-genYou can check out the example code in load-gen.ts. The configurable settings are at the bottom of the file.
If your application has traffic patterns that result in a high number of concurrent requests to Momento, it is possible that the nodejs event loop can become overwhelmed and cause the observed client-side latencies to rise drastically. Sometimes this results in client-side timeouts even when the server-side latencies are very low.
If your application may have a lot of duplicate requests, one solution is to de-duplicate the requests (aka request coalescing). With a tiny bit of code it's pretty easy to ensure that only one copy of each duplicate request actually goes out onto the network. This can reduce the amount of work that ends up in the node.js event loop, and dramatically improve the performance of your application.
This repo includes a request-coalescer, to allow you to experiment with performance in your environment based on different configurations.
The request-coalescing example shows difference between latencies when a basic load-generator is used v/s when request-coalescer is used.
Check out the configuration settings at the bottom of the 'request-coalescing.ts' to see how different configurations impact performance.
Here are some stats and steps on how to run the example:
Change the configuration found at the end of the file to
maxRequestsPerSecond = 1000 and numberOfConcurrentRequests = 1000
to achieve the stats shown below:
Stats should like:
| Request type, Metric | Before Coalescing | After Coalescing |
|---|---|---|
| set, p50 | 60 | 20 |
| set, p99 | 445 | 37 |
| get, p50 | 40 | 20 |
| get, p99 | 140 | 32 |
| set, % of requests coalesced | - | 49.5% |
| get, % of requests coalesced | - | 49.5% |
It's only about 50 lines of code to create a coalescing wrapper for the Momento cache client. You can see our implementation here: https://github.com/momentohq/client-sdk-nodejs/blob/main/examples/utils/momento-client-with-coalescing.ts
# Run example request coalescing
MOMENTO_API_KEY=<YOUR API KEY> npm run request-coalsecingYou can check out the example code in request-coalescing.ts. The configurable settings are at the bottom of the file.
If you have questions or need help experimenting further, please reach out to us!
For more info, visit our website at https://gomomento.com!