🧱 Databricks SQL & Serverless

SQL Warehouse, Databricks SQL Editor, Dashboards, Photon

Databricks Data Engineer · DE-Associate · ⏱️ 18 min · ❓ 3 savol

🏦 Databricks SQL nima?

Databricks SQL — BI va SQL analytics uchun maxsus muhit. SQL Warehouse (avval SQL Endpoint): compute resource SQL queries uchun. Serverless SQL Warehouse — compute Databricks Cloud da, siz node boshqarmaysiz, sekundlarda start. Classic SQL Warehouse — EC2/VM da, sekinroq start.

⚡ Photon Engine

C++ da yozilgan vectorized query engine — Spark DataFrame va SQL ni tezlashtiradi. Delta Lake, Parquet uchun optimalangan. Aggregation, join, sort operatsiyalari 2-8x tezroq. SQL Warehouse da avtomatik yoqilgan. Photon — extra narx qo'shadi lekin xarajat/performance ratio yaxshi.

📊 SQL Editor & Dashboards

SQL Editor — query yozish, vizualizatsiya, snippet. Dashboard — SQL queries natijasini chart, table, counter bilan ko'rsatish. Alert — query natijasi threshold ni oshganda email/Slack notification. Scheduled refresh — dashboard ma'lumotlarini avtomatik yangilash.

🔗 JDBC/ODBC & BI Tools

SQL Warehouse JDBC/ODBC endpoint — Tableau, Power BI, Looker, DBeaver bilan ulanish. Partner Connect — Databricks UI dan to'g'ridan-to'g'ri BI tool integratsiya. Connection string: server hostname, HTTP path, access token.

💡 Asosiy nuqtalar

SQL Warehouse: Databricks SQL uchun compute (Serverless > Classic)
Serverless SQL: sekundlarda start, siz node boshqarmaysiz
Photon: C++ vectorized engine, 2-8x tezroq aggregation
SQL Alert: query natijasi bo'yicha notification
JDBC/ODBC: Tableau, Power BI integratsiya

🎯 Imtihon maslahatlari

SQL Warehouse ≠ All-Purpose Cluster: SQL Warehouse = BI/SQL, Cluster = notebook/ML
Serverless SQL Warehouse — "instant start" va auto-stop, Classic — 2-5 min start
Photon — SQL va DataFrame API ni tezlashtiradi, streaming yoki ML uchun emas
Dashboard alert — query natijasi threshold ni oshganda notification
SQL Editor da `/* ... */` comment qo'llaniladigan variant va query history bor

⚠️ Ko'p adashadigan

SQL Warehouse va Interactive Cluster ni bir xil deb o'ylash — turli compute, turli use case
Photon barcha Spark operation uchun deb o'ylash — Photon SQL va DataFrame, ML/streaming cheklangan
Classic SQL Warehouse cold start vaqtini e'tiborsiz qoldirish — production da Serverless afzal

🧠 Eslab qolish: "SQL Warehouse = BI garaj": Serverless=elektr avtomobil (tez qo'zg'aladi), Classic=benzin (isitish kerak). "Photon = turbo dvigatel"