About this article: This article is reviewed against the cited sources. This is source research, not lab validation. Check the stated versions and your own requirements before making changes.
Not every disk on a hosted server needs the fastest storage. Database data, database logs and SQL Server's temporary database are the volumes that suffer when storage is slow, so they belong on the fast tier. File shares holding documents can often sit on a general tier, depending on how they are used, and archives on the cheapest capacity tier. Backup staging areas can also sit on a capacity tier, but only once measurements show that tier can take in each night's backup within the backup window and read it back fast enough to meet your restore-time target. Decide from measurements of each volume, not one figure for the whole server.
Why it matters
Fast storage costs more per gigabyte. A quote that places a 2 TB document archive on the same tier as a 100 GB database pays a premium for speed nobody uses. The opposite mistake, putting the database on slow storage to save money, shows up as slow screens for every user. Splitting volumes by need lets you spend where it changes the experience.
How it works
What tiers can mean
Providers name tiers differently; a three-level split might look like this:
- A performance tier: low delay per request and many requests per second, usually solid-state.
- A general tier: adequate for everyday file and application use.
- A capacity tier: lowest cost per gigabyte, slower, suited to data that is written once and read rarely.
Ask your provider for each tier's expected delay (latency) and request rate (IOPS), not only its name.
Which SQL Server files are sensitive
Every SQL Server database has at least two kinds of file. Microsoft explains that "Log files contain the information that is required to recover all transactions in the database". SQL Server's write-ahead logging "guarantees that no data modifications are written to disk before the associated log record is written to disk", so slow log writes slow down every change users save.
Microsoft's file placement recommendation is plain: "Don't put the transaction log files on the same physical disk that has the other files and filegroups." In hosting you rarely control the physical disks, and two virtual disks on the same shared storage pool may sit on exactly the same physical drives. A separate virtual disk keeps the log easy to size, monitor and place, but on its own it does not give the physical separation Microsoft describes. Ask the provider whether the log disk can sit on a different storage pool or datastore, or whether the tier guarantees a write delay under load that makes the separation matter less.
SQL Server's temporary workspace, tempdb, gets its own advice: "Put the tempdb database on a fast I/O subsystem." Busy reports and sorting lean on it heavily.
Measure before you choose
Microsoft's storage guidance for SQL Server on cloud virtual machines starts with measurement: "Monitor the application and determine storage bandwidth and latency requirements for SQL Server data, log, and tempdb files before choosing the disk type." That page's disk-type advice is specific to Azure, but the measure-first step applies to any hosted server: record each volume's request rate, throughput and delay over a busy period, then match each to a tier. Our published article Reading a storage specification: IOPS, throughput, latency and block size explains what those figures mean and how to compare them.
What to choose and why
| Volume | Suggested tier | Reason |
|---|---|---|
| Operating system | General | Modest, steady use |
| SQL Server data files | Performance | Reads and writes from every user action |
| SQL Server log files | Performance, separate disk (separate pool if offered) | Every saved change waits for the log write |
| tempdb | Performance | Microsoft advises fast I/O for tempdb |
| Application program files | General | Loaded once, then mostly cached |
| File shares (documents) | General | Opened one at a time by people; delay is less visible |
| Archive and old projects | Capacity | Written once, read rarely |
| Backup staging area | Capacity, if measured sustained write and read speeds meet the backup window and restore-time target | Large, sequential writes; sequential work can still be the bottleneck for a backup window or a restore, so measure before choosing |
The table is our editorial guidance built on the Microsoft sources above; measurement should override it. A file share used by a design team working on large files directly from the server may need a faster tier than this suggests.
A worked layout
For a hypothetical 30-person firm with one SQL-based application:
- Disk 1, general tier: Windows and the application.
- Disk 2, performance tier: SQL Server data files.
- Disk 3, performance tier: SQL Server log files.
- Disk 4, performance tier: tempdb (our suggestion: it could share disk 2 if measurements show little tempdb use).
- Disk 5, general tier: file shares.
- Disk 6, capacity tier: archive.
The example is illustrative.
Questions to ask your provider
- Which tiers are available, and what latency and IOPS does each deliver under load?
- Can different disks on one server sit on different tiers, or on different storage pools?
- Can a disk move between tiers later, and does that need downtime?
- Is pricing per gigabyte, per performance level or both?
Read next
Related articles you may want next.
- HostingReading a storage specification: IOPS, throughput, latency and block sizeHow the four figures in a hosting or cloud storage specification relate to each other, which one a business application depends on, how to measure what your server needs, and the questions to ask when comparing offers.Read article
- HostingSizing a business application server from measurements, not the vendor's minimumWhich Windows performance counters to collect on an existing server, for how long and at what interval, and how to turn the results into processor, memory and disk sizes for a hosted replacement.Read article
- HostingAdding CPU, memory or disk to a hosted server: which changes need a restartPlan CPU, memory and disk upgrades for hosted servers. Check which changes support live expansion, the conditions that apply and when a planned restart is needed.Read article
Risks and limits
- Tier names are not standards. "Premium" at one provider may be slower than "standard" at another; compare measured figures.
- Averages hide peaks. Measure at month-end or other busy periods.
- Extending database volumes needs a planned window. Our article Adding CPU, memory or disk to a hosted server: which changes need a restart covers Microsoft's backup-and-stop advice.
- Separate virtual disks are not separate drives. Without a separate pool, they share the same physical storage.
- Separate disks add management. More disks mean more to monitor and size.
Sources
Opened 15 November 2026.
- Microsoft Learn: Database files and filegroups, page shows last updated 11 December 2025.
- Microsoft Learn: SQL Server transaction log architecture and management guide, page shows last updated 21 September 2026.
- Microsoft Learn: tempdb database, page shows last updated 6 August 2026.
- Microsoft Learn: Storage: Performance best practices for SQL Server on Azure VMs, page shows last updated 23 April 2026.
Related articles: Reading a storage specification: IOPS, throughput, latency and block size; Sizing a business application server from measurements.
Related service: Cloud hosting.
Learning notes
- Database data, logs and tempdb need fast storage; archives do not.
- Microsoft advises keeping transaction logs off the physical disk holding the other database files; a separate virtual disk on the same pool is not the same thing.
- Measure each volume's delay and request rate before choosing tiers.
- Tier names differ by provider; compare measured figures.
This article was originally published on .