Tiered storage is a method of organizing storage resources that enables you to place data according to priority and performance requirements. It is typically used to reduce the cost of storage by reserving expensive high performance resources for high value data.
Tiered storage architectures incorporate different types of storage media or different classes of storage services. These types may support different data formats, different retrieval speeds, or different redundancy levels. For example, a typical storage tier architecture might use flash storage for the most frequently accessed data, hard disk drives for general use data, and cloud storage for archival storage.
A tiered storage architecture is a commonly used strategy in hybrid cloud and multicloud storage environments, and can be used both for public cloud and on-premise storage systems.
In this article, we will focus on the:
Tiered storage was first implemented by IBM in its mainframe computers. This implementation used multiple configurations of Serial Advanced Technology Attachment (SATA) and Serial-Attached SCSI (SAS) hard drives. These resources were arranged in a redundant array of independent disks (RAID) configuration and data blocks were written with striping and short stroking to increase performance.
This method required significant manual configuration and was beyond the scope of many smaller organizations capabilities. As tiered storage gained popularity, however, the process was simplified with hierarchical storage management (HSM) software.
HSM software enables organizations to automate much of the work that goes into storage tiering. For example, an HSM software can move data dynamically between RAID groups, drive types, or storage systems according to policies based on access history or metadata flags. This enables IT teams to focus on initial configurations while leaving maintenance and storage optimization to the software’s algorithms.
There are five storage tier types you can choose from when defining your tiered architecture. You do not have to use all five types nor do you have to use them sequentially.
Tier 1 storage is designed to store time-sensitive and highly volatile data that you need to access with minimal latency. It is based on flash memory optimized for performance. This tier is often used in environments where the revenue generated by the data is significantly greater than storage costs. For example, financial trading environments.
Tier 2 storage is designed to store customer-facing systems, high-performance applications, and transactional data with little latency. It is based on solid state drives (SSD) but does not have to be optimized for performance like tier 1 storage.
Tier 3 storage is designed to store data that you need to access frequently with reasonable delay, such as emails, customer resource management or enterprise resource planning data. It is based on medium to high performance hard disk drives (HDD). Most architectures have significantly more tier 3 storage than tier 1 or 2.
Tier 4 storage is designed to store infrequently accessed data that you want to remain easily accessible, such as older emails, recently completed transactions, or standard workload files. It may also be used to store metrics data or business information, such as that used for monthly or quarterly reports. It is typically based on SATA drives that prioritize capacity over performance.
Tier 5 storage is designed to store rarely or never accessed data, such as backups, archives, or compliance data. It is typically based on tape or optical storage media with long retrieval times. Several cloud providers, including the big three, also provide “cold” or tier 5 storage services with minimal capacity costs but higher retrieval costs.
Creating a tiered storage architecture on-premises or in a single cloud is possible but often doesn’t optimize costs. Some storage media are significantly easier to manage than others and variable workloads mean that purchased storage may sit idle at times. To avoid these issues and cost drains, some organizations choose to implement hybrid or multicloud storage.
In a hybrid storage system, primary (upper tier) storage is typically hosted on-premises or in a private cloud. This keeps high priority workloads close and firmly controlled. Meanwhile, lower tier storage can exist in the public cloud. This way you don’t have to worry about infrastructure management and can rely on the provider to ensure data redundancy and availability.
Related content: read our guide to hybrid cloud architecture
Multicloud systems can take the same approach as hybrid but leverage the services of multiple providers to ensure the best balance between cost and performance. Additionally, multicloud systems may be required if some applications require specialized components or storage structures.
Related content: read our guide to multicloud architecture
The falling cost of storage resources has increased the flexibility in storage architecture design that is available. In general, however, systems implement cloud storage to actively backup or create a failover for primary storage. Archive storage is also often in the cloud due to the long term, low cost, high durability options that are available. For example, Google Cloud Storage Nearline and AWS Glacier.
A major benefit of hybrid storage is that it enables you to move data between storage tiers more easily. Hybrid cloud services are linked or synced to on-premises resources, eliminating the need to move data manually. Instead you can use on-premises resources to cache your high priority data while the rest remains accessible with a short delay.
Another benefit of both hybrid and multicloud is that they make it easier to follow the backup best practice 3-2-1. This practice requires you to store three copies of data, in at least two locations, with at least one offsite. With cloud services, you can easily use different cloud services or regions to meet these requirements without establishing new data centers or purchasing duplicate resources.