Many techniques and approaches to protecting and maintaining access to data can be combined to support a resilient data storage infrastructure. RAID and mirroring can, by themselves, provide accessibility to data, however protecting data requires technologies such as backup, archiving, snapshot and point-in-time copies.
Today, demand for low-cost arrays is pushing RAID technology downward from the enterprise level, through workgroups and departments, and even to SMB and small office/home office (SOHO) applications.
RAID technology is at its best when you're not aware of it doing its job -- for example, if a disk drive fails and the only way you know of it is when you receive an alert. Given the large-capacity low-cost disk drives in use today, and availability of low-cost RAID adapter cards and storage systems, RAID is as relevant today as it was 20 years ago.
RAID can be implemented in firmware on hardware devices, or via software in operating systems, volume managers, virtualization appliances and applications. The different levels of RAID protection indicate options for data protection, availability and performance while reducing the number of disk drives needed.
The table below shows the eight RAID levels, their associated characteristics and the types of applications they can best benefit. Windows administrators may find themselves using differing RAID levels in your environment for different applications with different performance and availability needs. For instance, transaction data and other applications requiring high performance may be placed on RAID 0+1; RAID 1 will tend to be used for read-intensive data requiring high availability; and reference data may be placed on RAID-5, where it can be combined with write cache to offset write-penalty performance.
|RAID 0||Disk spreads data across two or more disk drives to improve I/O performance by performing parallel I/O. Each drive holds 1/nth of the data, where n is the number of disks.||Performance for reads and writes with no data redundancy. RAID 0 by itself should only be used for applications that can tolerate loss of access to data, and data that can reproduced from other sources.|
|RAID 1||Disk mirroring provides data protection and enhanced read performance. RAID-1 mirrors data across two or more disks so that each disk is identical to the other. RAID-1 utilizes n+n (1+1, 2+2, 3+3) protection, increasing the number of disks.||Read-intensive OLTP and transactional data for high performance and availability. Other applications that can benefit from RAID-1 include email, operating system, application files, and read and random intensive environments.|
|RAID 0+1||Stripe and mirroring of data to provide performance (stripe) and availability (mirroring) using n+n number of devices. Loss of a disk drive does not impact performance or availability, as would be the case with RAID 0.||OLTP and I/O-intensive applications requiring high performance and high availability. This includes transaction logs, journal files and database indices, where cost measurement is based on dollar per I/O compared to dollar per unit of storage.|
|RAID 1+0 (10)||Like RAID 0+1, RAID 1+0 mirrors and stripes data to provide performance (stripe) and high availability (mirroring) using n+n number of devices. The differences is striping groups of disks together and then mirroring the stripe groups.||OLTP and I/O-intensive applications requiring performance and high availability. These include transaction logs, journal files and database indices where the cost measurement is based on dollar per I/O compared to dollar per unit of storage.|
|RAID 3||Stripe with dedicated parity at the byte level. RAID 3 has a single dedicated disk drive that stores parity information using an n+1 approach in terms of the number of devices needed.||Provides performance for video imaging, geophysics, life science and sequential processing applications, but is not suited for concurrent I/O operations from multiple users or I/O streams.|
|RAID 4||Same as RAID-3, but with block-level parity protection.||Using read and write cache; well-suited to file serving environments.|
|RAID 5||Disk striping with rotating parity protection using n+1 components providing good availability read performance for concurrent users and I/O streams. Using a hot spare disk drive, data can be reconstructed (drive rebuild) to protect against a second failure once completed.||Reduces the number of components required while providing good availability, performance for reads. However, write performance will be affected if write cache not utilized. Applications include reference data, read-intensive database tables, general file sharing and Web applications.|
|RAID 6||Dual parity to protect and provide oviparity in the event of a double drive failure during disk rebuild.||Address prolonged rebuild times associated with large capacity SATA and Fibre Channel disk drives.|
Considerations with regard to RAID for Windows-based environments include:
- Do you need server-based hardware (i.e., an adapter) or an external RAID-based storage system?
- Which disk drives and storage interfaces (SAS, SATA or Fibre Channel) do you need?
- Given your application and service requirements, which RAID levels do you need?
- If you are using large-capacity disk drives, do you need RAID 6? Or is RAID 5 sufficient?
- For host RAID adapters, how many internal, external and combo ports do you need?
- Do you need PCI-Express or traditional PCI support for adapters?
- What additional functionality do you need, including proactive disk drive rebuild?
- What type of packaging or card slot footprint requirements do you need?
- Do you need to support IOPs or bandwidth, reads or writes?
There are many RAID packaging options, including storage systems, controller cards or host-server based RAID adapters. Vendors of RAID adapter cards include Adaptec, AMCC, ATTO, HP and LSI, but there are many other OEMs.
This is the first in a series of articles that take a look at protecting and securing data for Windows environments.
About the author: Greg Schulz is founder and senior analyst of the StorageIO Group, an independent storage analyst firm, and author of the book Resilient Storage Networks, published by Elsevier. Chapter 12 of the book covers RAID in depth. Chapter 2, which covers data storage fundamentals is available for download.
More information on this topic:
- Tip: SATA, ASICs give boost to RAID 6
- Topics: RAID
- RSS: Sign up for our RSS feed to receive expert advice every day.