Other RDBMS — SQLite, SQL Server & Oracle

অন্যান্য RDBMS — SQLite, SQL Server, Oracle

Read: ~45 min Intermediate 12 practice problems Live SQLite runner

1. Beyond Postgres and MySQL

The two giants of open-source RDBMS — Postgres and MySQL — share the spotlight, but you will absolutely meet three other engines in your career: SQLite (the most-deployed database in the world, sitting inside every smartphone, browser, IoT device), Microsoft SQL Server (every mid-to-large Bangladeshi enterprise running .NET / Microsoft stack), and Oracle Database (banks, telcos, and most legacy ERP systems including those of several Bangladeshi banks).

Postgres ও MySQL ছাড়াও আরও তিনটি RDBMS আপনি অবশ্যই দেখা পাবেন — SQLite (পৃথিবীর সবচেয়ে বেশি deploy হওয়া database, প্রতিটি smartphone-এ আছে), SQL Server (Microsoft / .NET-এর enterprise stack), এবং Oracle (ব্যাংক, telco, ERP — বাংলাদেশের অনেক ব্যাংকেই Oracle চলে)। এই module-এ প্রত্যেকটির বিশেষ দিক, কোথায় meet করবেন, এবং SQL syntax কীভাবে আলাদা হয় — সেটি দেখবো।

We'll close with a cross-DB portability cheat sheet — what's standard SQL (works everywhere) and what's vendor-specific — and a decision matrix to help you pick.

2. SQLite — The Most-Deployed Database on Earth

SQLite is not a server. It is a tiny C library — about 800 KB — that opens a file and exposes a SQL interface on top of it. No daemon, no port, no users, no backup utility. It is shipped inside applications. Every Android phone, every iPhone, every Chrome browser, every Firefox, every WhatsApp client, every Skype, every aircraft (Airbus uses it on the A350) and every Bangla Express bus ticket app you've ever used — they all carry SQLite quietly inside.

SQLite কোনো server নয় — এটি একটি ছোট C library। প্রতিটি Android phone, প্রতিটি browser, এমনকি অনেক embedded device-এ SQLite চলছে। যাঁদের কাজ এই page-এ চলছে — সেটিও আসলে browser-এর ভেতরে চালানো একটি SQLite। কোনো install, কোনো config নেই — শুধু একটি .db ফাইল।

2.1 Architecture & quirks

Property	SQLite	Why it matters
Deployment	One file (`app.db`) + one library.	No DBA, no service to crash.
Concurrency	One writer at a time; many readers.	Bad for multi-tenant servers; perfect for single-user apps.
Type system	"Type affinity" — values keep their declared affinity but can hold others.	You can put text into an INTEGER column. Surprising, mostly harmless.
Foreign keys	Off by default! Enable with `PRAGMA foreign_keys = ON;`	Easy footgun — referential integrity silently absent.
Transactions	Fully ACID. Default isolation: serializable.	Stronger than most servers' default.
Auto-increment	`INTEGER PRIMARY KEY` auto-increments via ROWID.	No `SERIAL` / `AUTO_INCREMENT` keyword needed.

2.2 WAL mode — for concurrent readers + writer

By default SQLite uses rollback-journal mode: a writer blocks all readers. Switching to WAL mode (write-ahead log) lets readers proceed in parallel with a writer — at the cost of an extra .db-wal file. For any app with concurrent access, turn this on.

sqlite_setup.sql

-- Recommended settings for any production-quality SQLite app
PRAGMA journal_mode = WAL;        -- readers + 1 writer concurrent
PRAGMA synchronous  = NORMAL;     -- fast and safe enough
PRAGMA foreign_keys = ON;        -- enforce FK constraints (off by default!)
PRAGMA temp_store   = MEMORY;     -- temp tables in RAM

-- Sanity check
SELECT sqlite_version() AS ver;

2.3 FTS5 — built-in full-text search

SQLite includes FTS5, a full-text search engine, as a "virtual table" module. For an offline-first app — say a Bangla recipe app with a search box — this is everything you need, no Elasticsearch, no extra service.

fts5_demo.sql

-- Create a full-text search virtual table
CREATE VIRTUAL TABLE recipe_fts USING fts5(title, body);

INSERT INTO recipe_fts(title, body) VALUES
 ('Beef Bhuna',    'A slow cooked spicy Bangladeshi beef curry with onions and ginger.'),
 ('Chicken Korma', 'A creamy mild dish often served at weddings.'),
 ('Hilsa Paturi',  'Hilsa fish wrapped in banana leaf with mustard paste.');

-- Search "beef curry" — FTS5 ranks results by relevance
SELECT title, body
FROM   recipe_fts
WHERE  recipe_fts MATCH 'beef curry'
ORDER BY rank;

FTS5-এ MATCH ব্যবহার করে শব্দ-ভিত্তিক search করা যায়; rank দিয়ে relevance অনুসারে sort। ছোট-মাঝারি app-এর জন্য Elasticsearch বা Solr-এর প্রয়োজন নেই।

2.4 R-tree — spatial / range indexes

SQLite ships an R-tree module — a spatial index for "find shapes that overlap a box." Useful for offline maps, asset bounding boxes, calendar overlap problems.

rtree_demo.sql

CREATE VIRTUAL TABLE building_idx USING rtree(
   id, min_lng, max_lng, min_lat, max_lat
);
INSERT INTO building_idx VALUES
 (1, 90.39, 90.41, 23.78, 23.80),  -- Banani area
 (2, 90.41, 90.42, 23.81, 23.82);  -- Gulshan

-- Buildings whose bounding box overlaps a given query box
SELECT id
FROM   building_idx
WHERE  max_lng >= 90.40 AND min_lng <= 90.415
  AND  max_lat >= 23.79 AND min_lat <= 23.80;

SQLite is the right answer when …data fits on one machine, you don't need cross-machine concurrency, and you want zero operations. Mobile apps, desktop tools, single-tenant SaaS prototypes, embedded devices, and CI test fixtures — all SQLite's home turf. The wrong answer when many writers must hit the same DB at once.

3. Microsoft SQL Server & T-SQL

SQL Server is Microsoft's flagship RDBMS, the default choice anywhere a .NET back-end runs. In Bangladesh you'll meet it at corporate banks, large garment ERPs, government MIS systems, and most organisations whose IT department is "Microsoft-shop."

SQL Server Microsoft-এর এন্টারপ্রাইজ database। বাংলাদেশে .NET stack-এ চলা প্রায় সব corporate ও government system-এ এটি দেখা যায়। SQL Server-এর SQL dialect-কে বলা হয় T-SQL — Transact-SQL, যা Sybase থেকে উত্তরাধিকার সূত্রে এসেছে।

3.1 T-SQL — the dialect

Concept	T-SQL syntax	Notes
Auto-increment	`id INT IDENTITY(1,1) PRIMARY KEY`	vs Postgres `SERIAL`, MySQL `AUTO_INCREMENT`, SQLite `INTEGER PRIMARY KEY`.
String concat	`'a' + 'b'`	Postgres / SQLite use `\|\|`; MySQL uses `CONCAT()`.
LIMIT	`SELECT TOP 10 * FROM t` or `OFFSET … FETCH NEXT 10 ROWS ONLY`	No bare `LIMIT` keyword.
Variables	`DECLARE @x INT = 5;`	The `@` sigil is mandatory.
IF / control flow	`IF … BEGIN … END`	No `END IF`.
Date now	`GETDATE()` or `SYSDATETIME()`	Postgres / MySQL use `NOW()` / `CURRENT_TIMESTAMP`.
Schema separator	`dbo.MyTable` (database.schema.table)	3-part naming common.

tsql.sql

-- Classic T-SQL stored procedure
CREATE PROCEDURE dbo.GetTopCustomers
    @MinSpend DECIMAL(12,2)
AS
BEGIN
    SET NOCOUNT ON;

    DECLARE @AsOf DATETIME2 = SYSUTCDATETIME();

    SELECT TOP 10
           c.customer_id,
           c.full_name,
           SUM(o.amount) AS total_spend,
           @AsOf         AS as_of
    FROM   dbo.[order]    AS o
    JOIN   dbo.customer   AS c ON c.customer_id = o.customer_id
    GROUP BY c.customer_id, c.full_name
    HAVING SUM(o.amount) >= @MinSpend
    ORDER BY total_spend DESC;
END;

EXEC dbo.GetTopCustomers @MinSpend = 50000;

3.2 Columnstore indexes — analytics inside the OLTP DB

A unique strength: SQL Server's columnstore indexes (clustered or non-clustered) store data column-by-column with heavy compression — so a single SQL Server can serve OLTP rows and aggregate billions of rows for dashboards. (Postgres has columnar via the citus_columnar / external extensions; MySQL has nothing comparable.)

tsql.sql

-- A columnstore index over a fact table; analytical queries fly on it
CREATE CLUSTERED COLUMNSTORE INDEX ccx_fact_sales
ON dbo.fact_sales;

-- Same SQL as before; the optimizer chooses the columnstore path
SELECT region, YEAR(order_date) AS yr, SUM(amount) AS revenue
FROM   dbo.fact_sales
GROUP BY region, YEAR(order_date);

3.3 Always On Availability Groups

HA in SQL Server is built on Availability Groups — a set of databases failing over together with synchronous or asynchronous replicas, automatic listener routing, and read-only secondaries. The mental model is similar to MySQL's Group Replication or Postgres's Patroni-managed cluster, but tightly integrated with Windows Server Failover Clustering.

3.4 Azure SQL — the cloud cousin

Azure SQL Database and Azure SQL Managed Instance are managed SQL Server offerings on Azure. The dialect is essentially T-SQL with a few unsupported on-prem-only features (cross-database queries on Azure SQL DB, certain CLR features). For a Bangladeshi enterprise migrating to cloud, Azure SQL is often the lowest-friction path off an on-prem SQL Server.

SQL Server licensing Standard / Enterprise editions are paid — pricing per core. SQL Server Express is free up to 10 GB / 1 GB RAM / 1 socket — fine for small apps and CI. SQL Server on Linux + Docker (Developer edition) is free for non-production use; that's how most engineers learn T-SQL today.

4. Oracle Database — The Enterprise Heavyweight

Oracle Database is the world's most-used enterprise RDBMS by revenue. Banks, telecoms (Grameenphone, Robi historically), air traffic control, large ERPs (Oracle E-Business Suite, PeopleSoft, JD Edwards) — all run on Oracle. Many Bangladeshi banks (e.g. several state-owned and tier-1 private banks) run their core banking on Oracle.

Oracle Database পৃথিবীর সবচেয়ে বহুল ব্যবহৃত enterprise RDBMS — revenue হিসাবে। বাংলাদেশের অনেক ব্যাংক, telco এবং সরকারি বড় system Oracle-এ চলে। এর procedural ভাষার নাম PL/SQL — Postgres-এর PL/pgSQL ও SQL Server-এর T-SQL-এর কাছাকাছি, কিন্তু আরও পুরোনো ও বেশি feature-সমৃদ্ধ।

4.1 PL/SQL — Oracle's procedural language

Feature	PL/SQL	Equivalent elsewhere
Block	`DECLARE … BEGIN … EXCEPTION … END;`	Postgres `DO $$ … $$;` / T-SQL `BEGIN … END`.
Sequences	`seq.NEXTVAL`	Postgres `nextval('seq')`.
Identity column	`id NUMBER GENERATED ALWAYS AS IDENTITY` (since 12c)	Same SQL standard as Postgres / SQL Server.
Concat / null	`NVL(x, 0)` = `COALESCE(x, 0)`	Both work in modern Oracle.
Pagination	`FETCH FIRST 10 ROWS ONLY` (12c+)	Standard SQL, also in PG / SQL Server.
Pseudo-row	`ROWNUM`, `ROWID`	Oracle-specific; older code uses these.
Dual table	`SELECT 1 FROM dual`	A famous one-row pseudo-table for `SELECT` without `FROM`.

plsql.sql

-- Oracle PL/SQL block: process pending orders, with exception handling
DECLARE
    v_id    orders.id%TYPE;
    v_amt   orders.amount%TYPE;
    n       PLS_INTEGER := 0;
BEGIN
    FOR rec IN (
        SELECT id, amount FROM orders WHERE status = 'PENDING'
    ) LOOP
        BEGIN
            UPDATE orders
               SET status = 'CONFIRMED', confirmed_at = SYSTIMESTAMP
             WHERE id = rec.id;
            n := n + 1;
        EXCEPTION
            WHEN OTHERS THEN
                INSERT INTO order_error(order_id, msg, ts)
                VALUES (rec.id, SQLERRM, SYSTIMESTAMP);
        END;
    END LOOP;
    DBMS_OUTPUT.PUT_LINE('Processed ' || n || ' orders.');
    COMMIT;
END;
/

4.2 AWR & Statspack — Oracle's performance brain

Oracle ships with the Automatic Workload Repository (AWR) — periodic snapshots of every stat the database knows: top SQL by CPU, by I/O, wait events, latch contention. An AWR report is what DBAs call up after any incident. Open-source Statspack is a free predecessor still bundled.

4.3 RAC — Real Application Clusters

Oracle RAC is a shared-storage active/active cluster: many instances on different nodes all reading and writing the same database files via a fast interconnect. It's how Oracle achieves "five-nines" availability for core banking. Operationally complex; license costs are notorious. Postgres and MySQL solve the same problem differently — typically with logical replication and external proxies, not shared storage.

Honest reality of Oracle Oracle is technically excellent — partitioning, parallelism, and recovery features were "Oracle-exclusive" for decades. The cost, however, is famously high (per-core licensing in the tens of thousands of USD) and the audit risk is real. New greenfield projects rarely pick Oracle in 2026; you'll meet it in existing systems and during migrations off Oracle (often onto Postgres).

5. Cross-DB SQL Portability — What's Standard, What Isn't

"Standard SQL" sounds reassuring, but every engine has its own dialect. The good news: a large portable subset exists. The bad news: small details bite when you write a query intended to run on multiple engines. Here is the cheat sheet.

Topic	SQLite	MySQL	PostgreSQL	SQL Server	Oracle
Auto-increment	`INTEGER PRIMARY KEY`	`AUTO_INCREMENT`	`SERIAL` / `GENERATED … IDENTITY`	`IDENTITY(1,1)`	`GENERATED … IDENTITY` (12c+) / sequence + trigger
String concat	`\|\|`	`CONCAT(a,b)` / `+` if mode	`\|\|`	`+` or `CONCAT()`	`\|\|`
Now()	`CURRENT_TIMESTAMP`	`NOW()` / `CURRENT_TIMESTAMP`	`NOW()` / `CURRENT_TIMESTAMP`	`GETDATE()` / `SYSDATETIME()`	`SYSDATE` / `SYSTIMESTAMP`
LIMIT / TOP	`LIMIT 10`	`LIMIT 10`	`LIMIT 10` / `FETCH FIRST 10`	`TOP 10` / `FETCH NEXT 10`	`FETCH FIRST 10 ROWS ONLY`
Coalesce NULL	`IFNULL` / `COALESCE`	`IFNULL` / `COALESCE`	`COALESCE`	`ISNULL` / `COALESCE`	`NVL` / `COALESCE`
Boolean	0 / 1 (no real BOOLEAN)	`BOOLEAN` = `TINYINT(1)`	Real `BOOLEAN`	`BIT`	No native; `NUMBER(1)` or `CHAR(1)`
Recursive CTE	`WITH RECURSIVE`	`WITH RECURSIVE` (8+)	`WITH RECURSIVE`	`WITH` (recursive marker implicit)	`WITH` (implicit) or `CONNECT BY` (Oracle-only legacy)
Quoting identifiers	`"name"` or `[name]`	`name`	`"name"`	`[name]` or `"name"`	`"name"`
Substring	`substr(s,1,3)`	`SUBSTRING(s,1,3)`	`substring(s from 1 for 3)`	`SUBSTRING(s,1,3)`	`SUBSTR(s,1,3)`
Date math	`date('now','+1 day')`	`DATE_ADD(d, INTERVAL 1 DAY)`	`d + INTERVAL '1 day'`	`DATEADD(day, 1, d)`	`d + 1` (DATE) / `d + INTERVAL '1' DAY`

মূল কথা: Window function, CTE (recursive বাদে), JOIN, GROUP BY, HAVING, CASE, COALESCE — এগুলো প্রায় সব engine-এ এক রকম। বদলায় auto-increment, string concat, date function, এবং pagination syntax — এই কয়টি জিনিস সাবধানে দেখলেই আপনার SQL ৯০% portable হয়ে যাবে।

5.1 A portable query — works on all five

portable.sql

-- This query is identical across SQLite, MySQL 8, Postgres, SQL Server, Oracle 12c+
SELECT city,
       COUNT(*)            AS n_orders,
       COALESCE(SUM(amount), 0) AS revenue
FROM     orders
GROUP BY city
HAVING   SUM(amount) > 25000
ORDER BY revenue DESC;

6. Decision Matrix — Which RDBMS to Choose?

Scenario	Best fit	Why
Mobile / desktop app, single user, offline first	SQLite	Zero ops, full ACID, FTS5 + R-tree built in.
WordPress / PHP / cPanel hosting	MySQL / MariaDB	Ecosystem default; every plugin assumes it.
Modern startup, mixed OLTP + JSON + geo + vectors	PostgreSQL	JSONB, PostGIS, pgvector — one DB does it all.
.NET enterprise app on Windows / Azure	SQL Server / Azure SQL	First-class tooling, T-SQL skills abundant locally.
Core banking, telco billing, large ERP	Oracle	RAC, partitioning maturity, regulatory comfort.
Embedded device (router, car, set-top box)	SQLite	800 KB library, runs on anything with a filesystem.
Read-heavy SaaS at huge scale	PostgreSQL with read replicas, or MySQL + Vitess	Either works; pick by team skills.
OLAP / data warehouse	Not in this list — pick BigQuery, Snowflake, ClickHouse, or DuckDB.	Row-store RDBMS struggle past a few TB of analytics.

Practical advice for Bangladeshi engineers Learn three: SQLite (every interview, every embedded job), Postgres (every modern startup, every cloud-native job), and one enterprise engine — SQL Server or Oracle — depending on the local employers you target. MySQL skills come for free along the way because so much overlaps.

7. Glossary (শব্দকোষ)

Term	Meaning	বাংলায়
Embedded database	Library linked into the app process; no separate server.	Application-এর ভেতরেই চলে — আলাদা server নেই।
WAL mode (SQLite)	Write-ahead log letting readers + 1 writer run in parallel.	Reader ও writer-এর সমান্তরাল চলার জন্য SQLite-এর mode।
FTS5	SQLite's built-in full-text search virtual table.	SQLite-এর ভেতরে full-text search engine।
T-SQL	SQL Server's procedural SQL dialect.	SQL Server-এর procedural SQL।
Columnstore index	Column-oriented compressed storage for analytics.	Analytics-এর জন্য column-অভিমুখী compressed storage।
Always On AG	SQL Server's HA replica-group feature.	SQL Server-এর HA replica গ্রুপ।
PL/SQL	Oracle's procedural language.	Oracle-এর procedural ভাষা।
AWR	Oracle's automatic performance snapshot system.	Oracle-এর automatic performance snapshot।
RAC	Oracle's shared-storage active/active cluster.	Oracle-এর shared-storage active/active cluster।
Standard SQL	The portable subset defined by ISO/ANSI.	ISO/ANSI কর্তৃক সংজ্ঞায়িত portable SQL।

8. Practice Problems

A mix of decision-style questions, dialect translation, and a few SQLite-runnable demos using FTS5 / R-tree.

কিছু matching, কিছু dialect-এর অনুবাদ, কিছু এই page-এ চালানো যাবে।

Translate this Postgres-only snippet to T-SQL: SELECT NOW(), 'Hi ' || name FROM customer LIMIT 5;

Postgres-এর snippet-টি T-SQL-এ রূপান্তর করুন।
✨ Show Answer
```
SELECT TOP 5 GETDATE() AS now_ts, 'Hi ' + name AS greeting
FROM   customer;
```
পরিবর্তিত: NOW() → GETDATE(); || → +; LIMIT 5 → TOP 5।
Translate the same query to Oracle 12c+.

একই query Oracle 12c+-এ লিখুন।
✨ Show Answer
```
SELECT SYSTIMESTAMP AS now_ts, 'Hi ' || name AS greeting
FROM   customer
FETCH FIRST 5 ROWS ONLY;
```
A senior dev says "we don't need Elasticsearch for our offline Bangla recipe app — SQLite already does it." Which feature are they referring to and how does it work in one sentence?

কোন feature-এর কথা বলছে এবং কীভাবে কাজ করে?

✨ Show Answer

Answer: FTS5 — a built-in virtual-table module that builds an inverted index of tokens at insert/update time and supports the MATCH operator with relevance ranking, all without leaving SQLite.
Run the FTS5 demo from §2.3 and modify it to find recipes containing the word "fish."

FTS5 demo-তে "fish" শব্দটি খুঁজুন।
✨ Show Answer
ans4.sql
```
SELECT title
FROM   recipe_fts
WHERE  recipe_fts MATCH 'fish'
ORDER BY rank;
```
Why must you set PRAGMA foreign_keys = ON; in every SQLite session you care about referential integrity in?

SQLite-এ foreign key চালু করতে কী করতে হয় এবং কেন?

✨ Show Answer

Answer: SQLite parses FK constraints but does not enforce them by default — for backward compatibility with very old SQLite databases. Each new connection must explicitly enable enforcement with PRAGMA foreign_keys = ON;. Forgetting this is a classic bug: schemas look correct, integrity is silently absent.
When would a columnstore index in SQL Server beat a regular B-tree index for a query?

SQL Server-এ columnstore কখন B-tree-কে হারায়?

✨ Show Answer

Answer: When the query scans a large fraction of a fact table and aggregates a few columns — e.g. SUM(amount) GROUP BY region across 200M rows. Columnstore reads only the needed columns, in compressed form, and runs vectorized aggregations. B-tree shines for point lookups; columnstore shines for analytics scans.
Spot the Oracle-only construct: SELECT 1 FROM dual;. What is dual and what's the equivalent in Postgres?

dual কী, এবং Postgres-এ সমতুল্য কী?

✨ Show Answer

Answer: dual is a built-in single-row, single-column table Oracle requires when there's no real FROM target (Oracle insists on a FROM clause). In Postgres / SQLite / SQL Server / MySQL you can simply write SELECT 1;.
Choose a database and justify in one sentence: a school management system for 30 schools across Bangladesh, run on cheap shared hosting, with non-technical owners.

৩০টি স্কুলের জন্য shared hosting-এ system — কোন DB?

✨ Show Answer

Answer: MySQL / MariaDB — every cPanel host has it preinstalled with phpMyAdmin, deployment is one click, and the operational skill bar is the lowest. Postgres would also work but is rarer on shared hosting plans.
Choose a database: a large Bangladeshi private bank's core banking system, regulator-audited, must support 24x7 with seconds of failover.

Core banking — কোন DB?

✨ Show Answer

Answer: Oracle with RAC + Data Guard. Regulators are familiar with it, the vendor relationship and audit trail exist, and RAC delivers true active/active HA on shared storage. (In greenfield, some banks now consider Postgres + Patroni, but core banking conservatism still favours Oracle.)
Choose a database: a desktop Bangla typing tutor that students will install on their own laptops, no internet required.

Desktop typing tutor — কোন DB?

✨ Show Answer

Answer: SQLite — bundled with the app, single .db file, zero install, full ACID, FTS5 if needed for lesson search.
Recursive CTE syntax — write a tiny CTE that lists numbers 1..5. Verify it runs in SQLite (and so equally in Postgres / MySQL 8 / SQL Server / Oracle 12c+).

১ থেকে ৫ পর্যন্ত সংখ্যা একটি recursive CTE দিয়ে তৈরি করুন।
✨ Show Answer
ans11.sql
```
WITH RECURSIVE nums(n) AS (
    SELECT 1
    UNION ALL
    SELECT n + 1 FROM nums WHERE n < 5
)
SELECT n FROM nums;
```
In Oracle, drop the keyword RECURSIVE — Oracle's WITH infers it.
In one paragraph, explain why "write portable SQL" is a goal worth caring about even if you'll only ever ship on one engine.

এক RDBMS-এই deploy হলেও portable SQL লেখা কেন গুরুত্বপূর্ণ?

✨ Show Answer

Answer: Because "ever" is longer than you think. Companies switch databases (M&A, cloud migration, cost cutting), test fixtures may run on SQLite while production is Postgres, and reporting tools sometimes target a different engine for analytics. SQL written in the standard subset survives all of these transitions; SQL written with vendor sugar everywhere becomes a re-write project. Portability is also a code-quality signal — you tend to lean on what's expressive (CTEs, window functions, COALESCE) rather than what's lock-in (Oracle's CONNECT BY, T-SQL's + string concat).

Summary — Module 43

SQLite is the most-deployed database in the world — an embedded library, perfect for single-user and offline apps, with FTS5 and R-tree built in. SQL Server rules the .NET / Microsoft enterprise stack with T-SQL, columnstore indexes for analytics, and Always On Availability Groups for HA. Oracle is the heavyweight of banks, telcos and big ERPs, with PL/SQL, AWR for performance forensics, and RAC for shared-storage HA — paid handsomely for its track record. Underneath, almost all five major engines speak a shared standard SQL subset (CTEs, window functions, joins, group by, COALESCE), and differ in small but real ways for auto-increment, string concatenation, date functions and pagination. Knowing which features are portable and which are vendor-specific is what lets you write SQL that survives migrations and outlives any single product choice.

SQLite — embedded, single-user / offline app-এর জন্য সেরা; FTS5 ও R-tree বিল্ট-ইন। SQL Server — Microsoft stack, T-SQL, columnstore, Always On। Oracle — bank ও telco-এর enterprise; PL/SQL, AWR, RAC, কিন্তু দাম বেশি। পাঁচ engine-ই standard SQL-এর একটি বড় অংশ share করে — যা portable, সেগুলোকেই অগ্রাধিকার দিন; vendor-specific sugar (Oracle CONNECT BY, T-SQL +) যথাসম্ভব এড়িয়ে চলুন।

Next Module → NoSQL primer — when relational isn't the right fit (document, key-value, graph, time-series).

Prev: MySQL Next: NoSQL Primer