SLA RATE LIMIT

Contents

• SLA Rate Limit
• Setup
  • Installation
  • Configuration
• SLA Document
  • Plans
  • Rates
  • Quotas

SLA Rate Limit

SLA Rate Limit is an npm package containing a rate limitter middleware that can be integrated inside OAS Tools Core Library in order to limit or input some delay to server requests based on the SLA4OAI Standard.

Setup

Installation

In order to start using SLA Rate Limit just install it through your preferred package manager, in case of NPM:

npm install @oas-tools/sla-rate-limit

Once installed, import SLARateLimit middleware and call OAS Tools' use() function before initialization:

import http from "http";
import express from "express";
import { use, initialize } from "@oas-tools/core";
import { SLARateLimit } from "@oas-tools/sla-rate-limit";

const app = express();

use(SLARateLimit, {/* Config object */}, 2);
initialize(app).then(() => {
  http.createServer(app).listen(serverPort, () => {
    /* callback */
  });
})

Notice the third parameter used in the use function. Since the rate limitting action should be performed before processing any request, but after validating any security token, the SLA Rate Limit middleware is inserted in the position 2 of the express chain.

Configuration

The configuration is set through the second parameter of the use function. The table below describes the possible configuration options currently supported by the middleware:

| Param | Type | Description | Default | |------------------- |:--------: |---------------------------------------------------------------------------------------|------------------ | | slaFile | String | absolute or relative URI to the SLA file | api/oas-sla.yaml | | requestIdentifier | String | Name used in the SLA to identify the requests metric | requests | | scheme | String | Security scheme containing the token with the plan the user is suscribed to | apikey |

SLA document

This rate limit middleware requires a service level agreement file, in which the declaration for rates and quotas are found. This file should be located by default at api/oas-sla.yaml, but this option can be overriden through configuration, as explained above.

The SLA document must follow the SLA4OAI specification in order to declare dynamic and static windows for requests in a standard way. The example below defines a dynamic window of 1 request per second and a static window of 3 requests per minute for different endpoints:

sla: 1.0.0
context:
  id: rate-limit-sample
  type: plans
  api:
    $ref: ./oas-doc.yaml
  provider: ISAGroup
metrics:
  requests:
  type: "int64"
  description: "Number of requests"
plans:
  base:
    rates:
      /api/v1/resources/1:
        get:
          requests:
          - max: 3
            period: second
    quotas:
      /api/v1/resources:
        get:
          requests:
          - max: 3
            period: minute

This way, when making multiple requests to /api/v1/resources/1, the requests will be delayed in order to meet the rate criteria, whereas when making more requests than specified in the quotas object, the server response code will be 429 since the quota limit has been exceeded.

Plans

As shown in the example in the previous section, the SLA document must contain plans in which the rates and quotas are defined. The SLA Rate Limit middleware receives a token that must contain a plan attribute (by default is base, as explained in configuration section). This way, multiple plans containing different rates can be declared, making the server suscribe to one or another based on configuration (restaring the server is required when changing a plan).

Rates

Rates are managed by the express-slow-down middleware. This middleware will input delay on the requests in order to meet the dynamic window specified under the rates object in the SLA Document. Rates can be defined along quotas for the same endpoints. This situation is handled by the SLA Rate Limit middleware through chaining express-slow-down and express-rate-limit middlewares before registering them for the corresponding endpoint inside the express chain.

Quotas

Quotas, on the other hand, are managed by the express-rate-limit middleware. This middleware will make the server respond a 429 Too Many Requests when the quota is surprassed within the specified static window.