In this podcast, Simon Johnson, data center practice lead at GlassHouse Technologies, discusses the benefits and limitations of MAID, as well as MAID levels and how they operate.
Check out some of the highlights of this discussion below or listen to the entire podcast for a full discussion about MAID.
Here are some highlights from the MAID podcast: What is MAID? MAID is a storage technology in which only the active disk drives and active reads are spinning at a given time. So it's essentially a SATA storage array that you would see in a data centre, but there's a reduced number of spinning disks at one time. MAID technology is aimed at persistent data and in order to be made compliant, the solution must have a maximum of 50% of the drives spinning at one time. If you look closer at some of the vendors that provide MAID solutions, you will even see some that push down to a maximum of 25% of disks spinning at any given time. How does MAID work? MAID works by either completely shutting disks off or spinning disks down, depending on the version of MAID that you have. Both MAID 1 and MAID 2 spin down or switch off disks that are idle or not having data currently accessed. The MAID 1 approach is to turn off disk drives and cache significant pair savings, but this will not deliver any flexibility. Because the practice has such a drastic negative effect on application performance, it's often not practical to implement MAID 1. The goal of MAID 2 is to deliver energy savings to the broadest possible set of applications without affecting their performance. MAID 2 introduces this flexibility though intelligent pair management, which can be seen in a number of MAID levels. What are the benefits of MAID? The motivation of a MAID architecture is to deliver a solution that can exploit the inexpensive side of disk technology to create a commercially viable enterprise-class storage solution. This is done to exhibit much of the access and performance data integrity characteristics of a disk array, with the economic characteristics of a tape library. MAID is best suited for persistent data that's not static and data that can explore the more aggressive levels of spin-down and disk shut off. This will result in reduced power consumption and is the reason why MAID is pushed as a vital part of any green data centre. MAID provides a far greater footprint density, allowing you to leverage inexpensive SATA disks on a ratio of 3:1 on top of a normal array. So for every one disk in a data storage array, you would have three in MAID. What are the limitations of MAID and how can you address them? The limitations of MAID, or the considerations you need to be aware of when introducing it into your tiered storage operations, revolve around your data classification practices. MAID 2 offers different levels of drive spin-down, and if you're really going to exploit the energy savings that come from MAID, you're going to have to understand the profiles of the data that you're going to house in there. You don't want to be mixing data types across MAID levels. You want to ensure that data with the same I/O profiles and access levels are kept on the same level. By doing this, you can ensure that you actually have a solution that does spin down and gives you effective data pathing.
Here are some highlights from the MAID podcast:
MAID is a storage technology in which only the active disk drives and active reads are spinning at a given time. So it's essentially a SATA storage array that you would see in a data centre, but there's a reduced number of spinning disks at one time. MAID technology is aimed at persistent data and in order to be made compliant, the solution must have a maximum of 50% of the drives spinning at one time. If you look closer at some of the vendors that provide MAID solutions, you will even see some that push down to a maximum of 25% of disks spinning at any given time.
How does MAID work?
MAID works by either completely shutting disks off or spinning disks down, depending on the version of MAID that you have. Both MAID 1 and MAID 2 spin down or switch off disks that are idle or not having data currently accessed. The MAID 1 approach is to turn off disk drives and cache significant pair savings, but this will not deliver any flexibility. Because the practice has such a drastic negative effect on application performance, it's often not practical to implement MAID 1.
The goal of MAID 2 is to deliver energy savings to the broadest possible set of applications without affecting their performance. MAID 2 introduces this flexibility though intelligent pair management, which can be seen in a number of MAID levels.
What are the benefits of MAID?
The motivation of a MAID architecture is to deliver a solution that can exploit the inexpensive side of disk technology to create a commercially viable enterprise-class storage solution. This is done to exhibit much of the access and performance data integrity characteristics of a disk array, with the economic characteristics of a tape library.
MAID is best suited for persistent data that's not static and data that can explore the more aggressive levels of spin-down and disk shut off. This will result in reduced power consumption and is the reason why MAID is pushed as a vital part of any green data centre. MAID provides a far greater footprint density, allowing you to leverage inexpensive SATA disks on a ratio of 3:1 on top of a normal array. So for every one disk in a data storage array, you would have three in MAID.
What are the limitations of MAID and how can you address them?
The limitations of MAID, or the considerations you need to be aware of when introducing it into your tiered storage operations, revolve around your data classification practices. MAID 2 offers different levels of drive spin-down, and if you're really going to exploit the energy savings that come from MAID, you're going to have to understand the profiles of the data that you're going to house in there. You don't want to be mixing data types across MAID levels. You want to ensure that data with the same I/O profiles and access levels are kept on the same level. By doing this, you can ensure that you actually have a solution that does spin down and gives you effective data pathing.
