Storage tiering automatically matches data to particular media types according to a variety of user-defined policies to ensure data resides on the most cost-effective disk. Most storage arrays now come with storage tiering as a standard feature, but how do you add storage tiering to an existing environment?
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
In this interview, SearchStorage.co.UK Bureau Chief Antony Adshead speaks with Darren Janaway, senior consultant with GlassHouse Technologies (UK), about definitions of storage tiering and the various ways it can be added to an existing storage infrastructure.
Listen to the podcast on storage tiering or read the transcript below.
Download for later:
- Internet Explorer: Right Click > Save Target As
- Firefox: Right Click > Save Link As
SearchStorage.co.UK: What is storage tiering, and why should you add it to your environment?
Janaway: Storage tiering is the practice of moving data from one type of storage to another based on patterns of access, performance requirements or even capacity constraints.
As an example of access patterns, let’s consider a batch of high-resolution images taken 10 years ago and only looked at twice in the last five years. It really doesn’t make sense for these images to consume gigabytes of expensive 15,000 rpm Fibre Channel storage. The characteristics of 15k disk make them far more suitable for frequently accessed or important data that needs to be read or written as quickly as possible.
Rarely accessed 10-year-old image data is far more suited to slower, high-capacity SATA storage, which is generally cheaper to deploy and probably has a lower total cost of ownership.
Conversely, an Oracle database accessed every second of the day and with extremely high performance requirements should probably sit on fast, 15,000 rpm SAS or Fibre Channel storage, not SATA.
Storage tiering systems can automatically migrate data from faster to slower disk or slower to faster disk based on established policies.
So, what is a storage tier, and how should you define one? Well, the definition of what constitutes a storage tier is subjective and should really be aligned with business requirements.
An organisation might notionally define tier 1 storage as secure, high-performance and fault-tolerant, suitable for mission-critical business applications. The definition for the type of storage that would satisfy these requirements would probably be only SSD, Fibre Channel or SAS disk that would fit the bill. This leaves the storage with particular physical characteristics that constitute a tier and is often how businesses think of storage tiers.
Now let’s look at the definition of storage tiers from a physical perspective. As a rule of thumb it is generally accepted that Fibre Channel or SAS disks spinning at 15,000 rpm constitute tier 1 storage. Tier 2 would probably also consist of Fibre Channel or SAS, running at 10,000 rpm. Finally, SATA disks running at 7,200 rpm with much lower I/O rates would probably be considered tier 3.
So, what about SSD? Taken at its simplest, SSD is a bunch of nonvolatile memory chips masquerading as disk. It’s relatively new in the storage arena, and it’s still comparatively expensive technology, but with its high I/O capability and submillisecond access time, it’s generally considered tier 0.
Fibre Channel and SAS disk technologies are not quite as fast but are still considered high-performance and generally very reliable.
SearchStorage.co.UK: What types of products are available that can add storage tiering to an existing environment?
Janaway: The vast majority of storage vendors now provide some sort of storage tiering solution. A number of vendors supply arrays that come with management systems that can move data between defined tiers according to user-defined policies. Such policies define minimum or maximum rates of I/O that trigger the movement of data to a different tier, normally at the sub-LUN level—ie, at [the level of multiple blocks] rather than the entire LUN.
Some systems operate at block level, moving blocks between different tiers automatically or at specific times to balance throughput or capacity or to iron out disk hotspots, where a particular disk is accessed more frequently than others, causing bottlenecks.
Other solutions … operate at file level in a NAS environment and can not only move infrequently accessed file data between different tiers of storage but can also direct files based on type. For example, .mpg, or movie, files can be moved to a particular storage tier, while other types, such as .jpg [files] or Excel documents, could be routed to a different tier.
While some older arrays don’t support data tiering, data on such arrays can still be tiered by virtualising the existing array behind a new one that supports storage virtualisation and offers tiering. Once virtualised, hosts can still access data on the old array via the new one; the older virtualised array is now subject to the same storage tiering software as data stored locally on the new array.
Other arrays, while not supporting virtualisation in the same way, can still offer tiering, and once data has been migrated, storage tiering software in the array can move data around.
Other solutions include host-based solutions, almost an appliance that will virtualise and manage heterogenous arrays in a similar way to those arrays that offers virtualisation. This appliance offers a tiering solution that can move data between different types of storage, based on user-defined policies.
Finally, there is volume management software [that can run] on a SAN-attached host using a namespace spanning multiple volumes consisting of different tiers of storage, where files are placed on or moved between underlying volumes based on user-defined policies with no disruption to applications and end users.