In the first article devoted to Storage Spaces we introduced you to Storage Spaces and explained what it is in general.
In this article, we will discuss in more details which aspects you should consider when creating your own Storage Space.
So, you have disks on which you are going to create a Storage Spaces pool.
In Storage Spaces, you can use completely different disks. This is positioned as one of the notable Storage Spaces advantages. In the third and last article, we will discuss practical aspects of combining the disks at hand into a Storage Spaces pool in more detail.
Thin-provisioned or fixed?
Once you decided how many pools you need and distributed the disks over them, it’s time to create spaces in the pools. At this stage, you have to decide which space you need – thin-provisioned or fixed-provisioned. As it was explained in the first article, a thin-provisioned space allows you to add disk space as the need arises while a fixed-provisioned space requires all the capacity to be available at the moment of space creation.
Create a thin-provisioned space if you need a flexible volume that you are going to expand as it runs out of space.
Choose a fixed-provisioned space if you know in advance what and how much data will be stored on this space and data is now available and ready to be written to the space.
- You can create both thin-provisioned and fixed spaces on the same pool.
- Once you decided what space allocation scheme you need for your space, you cannot change it any longer. That means you cannot convert a fixed space to a thin-provisioned and vice versa.
- Disk space allocation type used for a space and space layout are two independent space characteristics, for example, a thin-provisioned space can be a parity space, simple, and so on.
- With a thin-provisioned space, you should be aware that when the time comes to add disk space to the corresponding pool, you need to add as many disks as there are columns in the space. For example, for a 3-disk thin-provisioned parity space you need to add at least three disks rather than one.
As of Windows Server 2012 R2, there are five layouts available in Storage Spaces: simple, mirror, three-way mirror, parity, and double-parity.
Having just one disk you can create only simple layout which, naturally, does not provide redundancy meaning that you lose all the data if a disk fails.
Mirror layout requires minimum two disks and unlike simple layout, it is redundant. When dealing with mirror spaces you should keep in mind that disk space overhead equals the total space capacity, i.e. if you create a 50 TB thin-provisioned mirror space, actually, somehow or other you will need to provide 100 TB. Thus, mirror spaces are optimal for not so large data volumes. With large volume of data look at the parity layout.
Three-way mirror is the most expensive layout in terms of disk space overhead because three data copies are stored allowing a space to survive a double disk failure. You need minimum five disks to create a three-way mirror space. Since the same considerations as for a regular mirror are applied to this layout, it is suggested to use it when you need a high level of data protection. With large data volumes consider the double parity layout instead.
Parity layout requires three disks minimum and survives a single disk failure. A parity space is all good except performance as to write speed since built-in obligatory journaling is used. Journaling provides advantages in stability allowing to avoid RAID write hole but significantly decreases write speed. Choose a parity layout over a mirror to store large archives when everything comes down to the price of disk space spent to provide overhead in mirror space. You should keep in mind that with large number of disks the probability of simultaneous failure of two or more disk increases, although, frankly speaking, it is not that high. However, if you have more than five disks, consider the double parity layout.
Double parity (Microsoft calls it dual parity) requires a minimum of four disks and is recommended when you create a space of many disks since in this case data protection against simultaneous double disk failure comes to the forefront. As to write speed, double parity space is a little worse than a parity space.
- Once you created a space, you cannot change its layout. For example, if you created a parity space you cannot convert it to, say, a simple space.
- You can create spaces of different layouts on the same disk set (pool), i.e. there can be a mix of parity, simple, or other spaces on the disks simultaneously.
- When you create a pool, you can add hotspare disks to it. If a disk fails, the Storage Spaces driver uses the hotspare disks to rebuild the redundant spaces with no action on your side. Truly, even without hotspare disks the Storage Spaces driver initiates a rebuild once a disk fails and starts to copy the lost data to the free disk space of other pool disks given that there is enough free disk space within redundancy limitations (for example, two data copies in a mirror space must be on two different disks). Hotspare disks guarantee free disk space availability once the need arises.
Once you have chosen a disk space allocation option and layout type for a space, you need to decide what filesystem to use. In Windows Server editions the Storage Spaces utility offers you to choose between NTFS and ReFS while in Windows 8 inside the Storage Spaces utility only NTFS is available.
If you deal with a mirror or parity space, choose ReFS since it was designed especially to work in conjunction with Storage Spaces. For example, if some errors, say, in metadata are detected, Storage Spaces corrects them without any user intervention and the space continues to work as if nothing has happened.
- In practice, outside the Storage Spaces utility you can format a space to any filesystem. Storage Spaces represents a space to Windows as a virtual disk; therefore, you can work with this disk in any disk management software. For example, you may go to Windows Disk Management and format a disk using FAT or exFAT.
- If you plan to store a large number of files, choose ReFS since it will be faster than NTFS. When you store a large number of small files the overhead of ReFS is noticeably larger than the NTFS overhead; however, as the number of files increases, the speed becomes more important than overhead.
- If you have a simple space or some other space for which you know in advance that it will be used to store a small number of files, choose NTFS since it will be faster and has a smaller overhead as compared to ReFS.
This is the second article of a series devoted to Storage Spaces. In the third and final article, we will discuss whether it is worth to create a Storage Spaces server from hard drives you find lying around.
Mathieu: Thank you Elena for this second article in the series on Microsoft Storage Space. Make sure to check back soon for the last part of this series.
Everyone, make sure to leave your questions, comments and suggestions in the comments section below.
Author bio: Written by Elena Pakhomova of www.ReclaiMe.com specializing in data recovery solutions for various storage devices. Elena has significant expertise with storage systems, including home and enterprise NAS/RAID units.