ProxyWhirl DocsGuides
Performance Tuning
Cache tiers, strategy latency tradeoffs, connection pooling, and throughput optimization.
Cache tiers
from proxywhirl import CacheConfig, ProxyConfiguration
config = ProxyConfiguration(
cache=CacheConfig(
l1_size=100,
l2_size=1000,
l3_enabled=True,
ttl_seconds=3600,
encryption=True,
)
)- L1: hot in-memory set (100–500 entries), sub-millisecond lookups
- L2: disk-backed JSONL shard, survives restarts
- L3: optional Redis for multi-instance deployments
See Cache Architecture.
Strategy characteristics
| Strategy | CPU | Memory | Latency | Best for |
|---|---|---|---|---|
| RoundRobin | Low | Low | Low | Uniform load |
| Weighted | Medium | Low | Low | Heterogeneous pools |
| PerformanceBased | High | Medium | Variable | Latency-sensitive |
| Random | Low | Low | Low | Unpredictable patterns |
| CostAware | Medium | Low | Low | Budget caps |
Connection reuse
Reuse HTTP clients to avoid per-request TLS handshakes:
import httpx
async with httpx.AsyncClient(
limits=httpx.Limits(max_keepalive_connections=20, max_connections=40)
) as client:
for _ in range(1000):
proxy = await whirl.get_proxy()
await client.get(url, proxy=proxy.to_url())Monitoring
Use MetricsCollector and Grafana dashboards to track cache hit rate, selection latency, and circuit breaker state. Target >80% cache hit rate in steady state.
See FAQ and Operations.