{"canonical_url":"https://engineer.company/glossary/sql-server/","content_markdown":"# SQL Server\n\n## What it is\n\nSQL Server, which many teams write as MSSQL or MS SQL, is a relational database\nmanagement system: a server program that keeps data in tables, enforces the\nrules between them and answers the applications that talk to it in Transact-SQL,\nor T-SQL, Microsoft's dialect of SQL. A client connects to an instance, which is\none running copy of the engine, and does its work inside one of the databases\nthat instance hosts. On disk, each of those databases is at least two files, one\nfor data and one for its transaction log.\n\nThe core component is the Database Engine, the service that stores, processes\nand secures data. It installs on Windows or Linux, in a Linux container or on a\nvirtual machine. Linux support began with SQL Server 2017 (14.x), using the same\nDatabase Engine as on Windows, and Red Hat Enterprise Linux and Ubuntu are\nsupported for every release since. The same engine also sits underneath several\nof Microsoft's hosted services, among them Azure SQL Database, Azure SQL Managed\nInstance and SQL database in Microsoft Fabric.\n\nReleases are named by year and numbered internally: 2017 is 14.x, 2019 is 15.x,\n2022 is 16.x, and SQL Server 2025, whose support began on 18 November 2025, is\n17.x. Around the engine sit further services. Integration Services (SSIS) builds\npackages for extract, transform and load work aimed at data warehousing, and SQL\nServer Agent runs jobs on a schedule. Starting with SQL Server 2025, on-premises\nreporting is consolidated under Power BI Report Server, and Data Quality\nServices and Master Data Services leave the product; 2022 and earlier releases\nkeep them.\n\nThe editions draw the commercial lines. For SQL Server 2022 they are Enterprise,\nStandard, Web, Express (free), Developer and Evaluation; the last two carry the\nfull Enterprise feature set, Developer licensed for development and test systems\nonly and Evaluation for 180 days. SQL Server 2025 redraws the list: Web edition\nends with 2022, Developer splits into Enterprise Developer and Standard\nDeveloper, and Express grows.\n\n## Why it matters\n\nA version and an edition together decide how large a server may get, which\nfeatures it has and how long Microsoft keeps fixing it.\n\nSize and features come from the edition. Express in SQL Server 2022 uses the\nlesser of one socket or four cores, at most 1,410 MB of memory for its buffer\npool, and holds at most 10 GB in each database. In SQL Server 2025 the Express\ncap rises to 50 GB, with the same compute and memory limits. Standard in 2022\nuses the lesser of four sockets or 24 cores and 128 GB of buffer pool; in 2025\nthose become 32 cores and 256 GB. For Enterprise, the edition table lists the\noperating system maximum for compute and buffer-pool memory under core-based\nlicensing. The legacy Enterprise Server+CAL model, unavailable for new\nagreements, limits each instance to 20 cores. SQL Server Agent comes with\nEnterprise, Standard and, in 2022, Web; Express leaves it out, so a free\ninstance that needs nightly backups needs a scheduler from outside the product.\n\nThe fix schedule comes from the version. Under Microsoft's Fixed Lifecycle\nPolicy each release gets at least ten years of support: five of mainstream\nsupport, with functional, performance, scalability and security updates, and\nfive of extended support, with security updates alone. For SQL Server 2019,\nmainstream support ran from November 2019 to the end of February 2025, and\nextended support continues to January 2030. Microsoft's update list shows CU32,\nfrom February 2025, as the latest cumulative update for 2019, with security\nreleases published on top of it since. SQL Server 2022 stays in mainstream\nsupport until January 2028 and in extended support until January 2033; SQL\nServer 2025 moves to extended support in January 2031 and leaves it in\nJanuary 2036.\n\nThe product also reaches people who never run it in production, because data\ntravels in its formats. A `.bak` file is a SQL Server backup, and reading one\nmeans restoring it into an instance, as one of the projects below did.\n\n## What it looks like in practice\n\nFor a mid-sized tech company, a DevOps role ran MS SQL Server 2019 alongside\nPostgreSQL 14 on\n[ten database servers built on Ubuntu 20.04 LTS VPS instances](/portfolio/ran-10-postgresql-and-ms-sql-servers-on-ubuntu/).\nBackups used `pg_dump` for PostgreSQL and SQL Server Agent jobs for MS SQL,\nwith retention policies and offsite storage. Prometheus and Grafana tracked the\nservers, and the result was 99.9% uptime across all ten.\n\nElsewhere, SQL Server was the format the data arrived in. Data for internal\ndashboards and client-facing Python APIs came from a remote warehouse behind an\nIPsec VPN, as compressed archives holding Parquet, CSV and `.bak` files.\n[The pipeline restored each backup into a local MSSQL Server instance with `RESTORE DATABASE`](/portfolio/designed-and-ran-6-etl-elt-pipelines/)\nand validated schema integrity. Staging loads used `bcp`, `psql` and SSIS\ndepending on the source format, and T-SQL and PL/pgSQL procedures cleaned and\nenriched the rows before they went on to BigQuery. With the `.bak` restores\nscripted along with the rest, the six pipelines did in under 20 minutes what\nhad taken 3+ hours by hand, and the data synced daily instead of weekly.\n\nA geographical map application, by contrast, needed to leave. It sat on a\nlegacy MSSQL database, and the team\n[moved it to PostgreSQL with the PostGIS extension for native geospatial indexing](/portfolio/improved-gis-map-performance-10x-on-postgresql/).\nMore than 50 million spatial records went through ETL into EPSG:4326, and GiST\nand SP-GiST indexes went onto the geometry and geography columns. Off SQL\nServer, the average spatial query took under 250 milliseconds instead of 2.5\nseconds, and backend CPU load fell by 65%.\n\n## Where it goes wrong\n\nEach of these starts at a boundary: between versions, between machines, or\nbetween a backup and its first use.\n\n**A backup travels forward in version, and only forward.** A backup restores\nonto the version of SQL Server that wrote it or onto a newer one. A database\nrestored from an earlier version is upgraded automatically on the way in, and\nthat upgrade is one-way. A `.bak` file written by a 2022 server therefore needs\na 2022 or 2025 instance, and a 2019 instance turns it away. Check which version\nwrote a file before building the instance that reads it.\n\n**The restore asks for folders the new server lacks.** A backup records where\nits data and log files lived on the source server, and each SQL Server version\nuses its own default path. On a different machine, list the logical file names\nfirst and then place each one with `MOVE`:\n\n```sql\nRESTORE FILELISTONLY\nFROM DISK = '/srv/incoming/sales.bak';\n\nRESTORE DATABASE Sales\nFROM DISK = '/srv/incoming/sales.bak'\nWITH MOVE 'Sales_Data' TO '/srv/mssql/data/Sales.mdf',\n     MOVE 'Sales_Log' TO '/srv/mssql/data/Sales_log.ldf';\n```\n\n**Backups pile up and nobody reads one back.** Microsoft's own\n[page on the subject](https://learn.microsoft.com/en-us/sql/relational-databases/backup-restore/back-up-and-restore-of-sql-server-databases)\nputs it bluntly: “You don't have a restore strategy until you test your\nbackups.” Restore a copy onto a test system on a schedule and run `DBCC CHECKDB`\nagainst it. Backup checksums detect problems with the backup media itself.\n\n**A bulk-copy file arrives without its table.** Files written by `bcp` carry the\nrows and nothing about their shape: no schema and no format information. When\nthe source table has changed or gone, importing them again needs either a format\nfile or a table with an identical definition. Microsoft recommends native format\n(`-n`) between two SQL Server instances and character format (`-c`) when the\ndata crosses into another system.\n\n**The free edition fills up.** Express suits learning and small applications,\nand 10 GB per database in 2022, or 50 GB in 2025, is a small allowance for\nanything that keeps history. Plan the move to a licensed edition, and its\nbudget, before the database reaches the cap.\n\n**An old release keeps running on the security-only track.** A server on SQL\nServer 2019 still works, and until January 2030 it receives security updates\nalone. The instances in the first project above ran 2019 between 2022 and 2024,\ninside mainstream support. Because backups restore forward, the way off such a\nrelease is a restore onto a newer instance, then a test of everything that talks\nto it.\n\n## Further reading\n\n- [What is SQL Server?](https://learn.microsoft.com/en-us/sql/sql-server/what-is-sql-server)\n  — Microsoft's overview of the engine, where it can be installed and the\n  services around it.\n- [SQL Server 2019 lifecycle](https://learn.microsoft.com/en-us/lifecycle/products/sql-server-2019)\n  — the start, mainstream and extended support dates for the release the\n  practice section ran.\n- [Editions and supported features of SQL Server 2025](https://learn.microsoft.com/en-us/sql/sql-server/editions-and-components-of-sql-server-2025)\n  — compute, memory and database-size limits per edition, with notes on what\n  changed since 2022.\n- [What is SQL Server on Linux?](https://learn.microsoft.com/en-us/sql/linux/sql-server-linux-overview)\n  — the release Linux support started with and the distributions each release\n  runs on.\n- [RESTORE statements (Transact-SQL)](https://learn.microsoft.com/en-us/sql/t-sql/statements/restore-statements-transact-sql)\n  — the full syntax, `MOVE` and `FILELISTONLY`, and the rules for restoring\n  across versions.\n- [bcp utility](https://learn.microsoft.com/en-us/sql/tools/bcp-utility) — bulk\n  copy in and out of the Database Engine, data formats and format files.\n\n\u003chttps://engineer.company/glossary/sql-server/\u003e\n","date_modified":"2026-11-09T21:50:25Z","description":"What SQL Server is, how its editions and support dates shape a deployment, and where restores, bulk loads and ageing versions catch teams out.","formats":{"gemtext":"https://engineer.company/glossary/sql-server/index.gmi","gopher":"https://engineer.company/glossary/sql-server/gophermap","html":"https://engineer.company/glossary/sql-server/","json":"https://engineer.company/glossary/sql-server/index.json","markdown":"https://engineer.company/glossary/sql-server/index.md","pdf":"https://engineer.company/glossary/sql-server/index.pdf"},"kind":"page","language":"en","title":"What is SQL Server?","url":"https://engineer.company/glossary/sql-server/","version":1}