High-Throughput Global Data Architecture:
1. Millions of Mobile Apps / IoT Devices ──► Write to ScyllaDB (Sub-0.4ms write latency worldwide!)
2. [ ScyllaDB Multi-Datacenter CDC Stream ]: Captures every row mutation asynchronously.
3. [ Apache Kafka / Flink ]: Buffers & batches CDC event streams.
4. ──► [ ClickHouse OLAP Cluster ]: Ingests batches to power real-time business intelligence dashboards! ✅

┌──────────────────┬───────────────────────────────┬───────────────────────────────┐
│ Dimension        │ ClickHouse (v26+)             │ ScyllaDB (v6.0+)              │
├──────────────────┼───────────────────────────────┼───────────────────────────────┤
│ Core Engine      │ Vectorized SIMD Columnar Merge│ **Seastar Share-Nothing       │
│                  │ Tree Storage Engine           │ Asynchronous Thread-per-Core**│
├──────────────────┼───────────────────────────────┼───────────────────────────────┤
│ Primary Workload │ **OLAP (Analytics & Scans)**  │ **OLTP / Fast Key-Value (Point│
│                  │ (Millions of rows scanned)    │ Reads & Single Row Writes)**  │
├──────────────────┼───────────────────────────────┼───────────────────────────────┤
│ Write Latency    │ 10 - 50 ms (Batch buffer)     │ **0.3 - 0.8 ms (Sub-ms p99!)**│
│ (Single Point)   │ (Prefers bulk inserts)        │ (Single-row real-time writes) │
├──────────────────┼───────────────────────────────┼───────────────────────────────┤
│ Multi-Datacenter │ ClickHouse Keeper / Raft      │ **Multi-DC Gossip / DynamoDB  │
│ Consensus Model  │ partition replication         │ Alternator & Raft Consensus** │
├──────────────────┼───────────────────────────────┼───────────────────────────────┤
│ CDC Stream       │ Kafka Engine / Materialized   │ **Native Built-In CDC Log     │
│ Capabilities     │ Views                         │ Tables per Partition**        │
└──────────────────┴───────────────────────────────┴───────────────────────────────┘

[ CPU Core 0 ] ──► Owns Partition Keys [0x0000 - 0x3FFF] ──► Dedicated In-Memory Cache & NVMe Queue
[ CPU Core 1 ] ──► Owns Partition Keys [0x4000 - 0x7FFF] ──► Dedicated In-Memory Cache & NVMe Queue
[ CPU Core 2 ] ──► Owns Partition Keys [0x8000 - 0xBFFF] ──► Dedicated In-Memory Cache & NVMe Queue
(Zero Mutex Locks! Zero Cross-Core Cache Line Invalidation!) ✅

-- ScyllaDB CQL: Enable Native CDC with 24-Hour Retention
CREATE TABLE user_wallet_transactions (
    user_id UUID,
    transaction_id TIMEUUID,
    amount_cents BIGINT,
    currency VARCHAR,
    created_at TIMESTAMP,
    PRIMARY KEY (user_id, transaction_id)
) WITH cdc = {'enabled': 'true', 'ttl': 86400};

# cdc_forwarder.py - Streaming ScyllaDB Mutations to ClickHouse
from cassandra.cluster import Cluster
import clickhouse_connect

# 1. Connect to ScyllaDB Cluster
scylla = Cluster(["10.0.1.10", "10.0.1.11"]).connect("production")

# 2. Connect to ClickHouse Cluster
ch_client = clickhouse_connect.get_client(host="clickhouse.mojostudio.in", username="default", password="secret_password")

def stream_wallet_cdc_events():
    # Read delta log from ScyllaDB CDC Table
    cdc_query = "SELECT * FROM user_wallet_transactions_scylla_cdc WHERE \"cdc$time\" > minTimeuuid(?) ALLOW FILTERING"
    rows = scylla.execute(cdc_query, [last_checkpoint_timestamp])
    
    batch = []
    for r in rows:
        batch.append([r.user_id, r.amount_cents, r.currency, r.created_at])
    
    # 3. High-Speed Columnar Batch Insert into ClickHouse
    if batch:
        ch_client.insert("production.wallet_analytics", batch, column_names=["user_id", "amount_cents", "currency", "created_at"])
        print(f"🚀 Streamed {len(batch)} CDC events to ClickHouse!")

Single Point Write Latency (Milliseconds - Lower is Better):
┌─────────────────────────────────────────────────────────┐
│ ClickHouse (Single Write): ████████████████████ 24.0 ms │
│ ScyllaDB Seastar:          █ 0.42 ms (57x Faster!) 🏆   │
└─────────────────────────────────────────────────────────┘