// PLATFORM / INFRASTRUCTURE_SPEC

THE ENGINE BEHIND SUB-MS QUOTAS

Limiter.io sits at your edge network and executes rate evaluation globally in less than a millisecond. By utilizing co-located memory segments and precompiled Redis Lua scripts, we guarantee complete concurrency safety.

01. Atomic Execution PipelineSPEC_V4.9.0

Multi-Tenant Isolation

Tokens and quotas are partitioned using cryptographically isolated namespaces. Tenant metrics never bleed across boundaries, guaranteeing safety.

Precompiled Lua Evaluation

Algorithm scripts are pre-loaded in memory using SHA hashes. Evaluated directly in Redis to prevent state discrepancies and race conditions.

Edge Proxy Fallback

Failures are bypassed gracefully. The client SDK features automatic fail-open strategies, maintaining API availability if backend clusters undergo updates.

// 02. ALGORITHM_STANDARDS

Token Bucket

ACTIVE

Maintains a rolling counter of available tokens refilled at a constant rate. Supports instantaneous bursts without throttling.

RECOMMENDED FOR: Standard API endpoints, user login routes, payment gates.

local key = KEYS[1]
local limit = tonumber(ARGV[1])
local rate = tonumber(ARGV[2])
local now = tonumber(ARGV[3])

Fixed Window

ACTIVE

Divides time into static windows (e.g. 1 minute) and tracks absolute request counts within that period. Discards window count upon boundary overlap.

RECOMMENDED FOR: Daily scraping limits, monthly data sync boundaries.

local count = redis.call("INCR", key)
if count == 1 then
  redis.call("EXPIRE", key, window)
end

Sliding Window Counter

ACTIVE

Uses a weighted average of the current and previous windows to compute the current rate, smoothing out boundary-crossing spikes.

RECOMMENDED FOR: High-traffic web hooks, global ingress protection.

local prev_count = redis.call("GET", prev_key) or 0
local curr_count = redis.call("GET", curr_key) or 0
local weight = (window_sec - elapsed) / window_sec

Sliding Window Log

ACTIVE

Logs every individual request timestamp in a Redis sorted set (ZSET). Evicts timestamps older than the window, offering complete accuracy.

RECOMMENDED FOR: High-value financial transfers, sensitive auth validation.

redis.call("ZREMRANGEBYSCORE", key, 0, min_score)
local current_requests = redis.call("ZCARD", key)
redis.call("ZADD", key, now, request_id)

Leaky Bucket

ACTIVE

Queues requests in a buffer that drips at a constant rate, smoothing out bursty traffic and enforcing a steady, strict output flow.

RECOMMENDED FOR: External third-party API sync, batch process ingestion.

local last_update = redis.call("HGET", key, "last")
local water = redis.call("HGET", key, "water")
local leaked = (now - last_update) * drip_rate

// 03. EDGE_LATENCY_METRICS

Avg Resolution

0.24ms

Cache Hit Rate

99.98%

Cluster Capacity

10M rps

Data Storage

SSD NVMe