brightdata-performance-tuning

# Bright Data Performance Tuning

## Overview

Optimize Bright Data scraping performance through connection pooling, response caching, concurrent request tuning, and smart product selection. Web Unlocker latency is typically 5-30s due to CAPTCHA solving; Scraping Browser sessions are 10-60s.

## Prerequisites

- Bright Data zone configured
- Understanding of async patterns
- Redis or file cache available (optional)

## Latency Benchmarks

| Product | P50 | P95 | P99 | Notes |
|---------|-----|-----|-----|-------|
| Web Unlocker (simple) | 3s | 8s | 15s | No CAPTCHA |
| Web Unlocker (CAPTCHA) | 10s | 25s | 45s | With CAPTCHA solving |
| Scraping Browser | 8s | 20s | 40s | Full browser render |
| SERP API (sync) | 2s | 5s | 10s | Search results |
| Residential Proxy | 1s | 3s | 8s | Raw proxy, no unblocking |

## Instructions

### Step 1: Choose the Right Product

```typescript
// Product selection matrix
function selectProduct(target: { js: boolean; captcha: boolean; structured: boolean }) {
  if (target.structured) return 'serp_api';       // Pre-parsed JSON
  if (!target.js && !target.captcha) return 'residential'; // Fastest
  if (target.js) return 'scraping_browser';         // Browser rendering
  return 'web_unlocker';                            // Best default
}
```

### Step 2: Connection Pooling with Keep-Alive

```typescript
import { Agent } from 'https';
import axios from 'axios';

// Reuse TCP connections to brd.superproxy.io
const httpsAgent = new Agent({
  keepAlive: true,
  maxSockets: 25,        // Match your concurrency limit
  maxFreeSockets: 5,
  timeout: 120000,
  rejectUnauthorized: false,
});

const client = axios.create({
  proxy: { host: 'brd.superproxy.io', port: 33335, auth: { username: proxyUser, password: proxyPass } },
  httpsAgent,
  timeout: 60000,
});
```

### Step 3: Response Caching Layer

```typescript
// src/brightdata/cache.ts — avoid re-scraping identical URLs
import { createHash } from 'crypto';
import { LRUCache } from 'lru-cache';

const memoryCache = new LRUCache<string, string>({
  max: 500,             // Max cached pages
  maxSize: 100_000_000, // 100MB total
  sizeCalculation: (v) => Buffer.byteLength(v),
  ttl: 3600000,         // 1 hour
});

export async function cachedScrape(
  url: string,
  scraper: (url: string) => Promise<string>,
  ttlMs?: number
): Promise<string> {
  const key = createHash('sha256').update(url).digest('hex');
  const cached = memoryCache.get(key);
  if (cached) {
    console.log(`Cache HIT: ${url}`);
    return cached;
  }

  const html = await scraper(url);
  memoryCache.set(key, html, { ttl: ttlMs });
  console.log(`Cache MISS: ${url} (${Buffer.byteLength(html)} bytes)`);
  return html;
}
```

### Step 4: Concurrent Scraping with Backpressure

```typescript
import PQueue from 'p-queue';

// Tune concurrency based on your plan and target site
const scrapeQueue = new PQueue({
  concurrency: 10,      // Concurrent proxy connections
  interval: 1000,       // Per second window
  intervalCap: 15,      // Max new requests per second
});

async function scrapeMany(urls: string[]): Promise<Map<string, string>> {
  const results = new Map<string, string>();

  await Promise.allSettled(
    urls.map(url =>
      scrapeQueue.add(async () => {
        const html = await cachedScrape(url, (u) => client.get(u).then(r => r.data));
        results.set(url, html);
      })
    )
  );

  console.log(`Scraped ${results.size}/${urls.length} successfully`);
  return results;
}
```

### Step 5: Use Async API for Bulk Jobs

For 100+ URLs, use the Web Scraper API instead of individual proxy requests:

```typescript
// Bulk collection — one API call, Bright Data handles parallelism
async function bulkScrape(urls: string[]) {
  const response = await fetch(
    `https://api.brightdata.com/datasets/v3/trigger?dataset_id=${DATASET_ID}&format=json`,
    {
      method: 'POST',
      headers: {
        'Authorization': `Bearer ${process.env.BRIGHTDATA_API_TOKEN}`,
        'Content-Type': 'application/json',
      },
      body: JSON.stringify(urls.map(url => ({ url }))),
    }
  );
  return response.json(); // Returns snapshot_id for status polling
}
// 1000 URLs via one trigger vs 1000 individual proxy requests
```

### Step 6: Performance Monitoring

```typescript
class ScrapeMetrics {
  private timings: number[] = [];
  private errors = 0;
  private cacheHits = 0;

  record(durationMs: number) { this.timings.push(durationMs); }
  recordError() { this.errors++; }
  recordCacheHit() { this.cacheHits++; }

  report() {
    const sorted = [...this.timings].sort((a, b) => a - b);
    return {
      count: sorted.length,
      errors: this.errors,
      cacheHits: this.cacheHits,
      p50: sorted[Math.floor(sorted.length * 0.5)] || 0,
      p95: sorted[Math.floor(sorted.length * 0.95)] || 0,
      p99: sorted[Math.floor(sorted.length * 0.99)] || 0,
    };
  }
}
```

## Output

- Right product selection per use case
- Connection pooling reducing TCP overhead
- Response cache avoiding duplicate scrapes
- Concurrent scraping with backpressure control
- Bulk API for large-scale jobs

## Error Handling

| Issue | Cause | Solution |
|-------|-------|----------|
| Slow scrapes | CAPTCHA solving overhead | Expected for Web Unlocker; use cache |
| Connection exhausted | Too many concurrent | Reduce p-queue concurrency |
| Memory pressure | Large cached pages | Set maxSize on LRU cache |
| Timeout storms | All requests hitting slow site | Add circuit breaker |

## Resources

- [Bright Data Products](https://brightdata.com/products)
- [Web Scraper API](https://docs.brightdata.com/scraping-automation/web-data-apis/web-scraper-api/overview)
- [p-queue](https://github.com/sindresorhus/p-queue)

## Next Steps

For cost optimization, see `brightdata-cost-tuning`.