Tape: It's not just for Tier 3 Storage Anymore
There’s a growing trend amongst the storage community to arbitrarily label tape as tier 3 storage. While the use of tape is rising rapidly, it is not fair to identify it for tier 3 use only. The idea of storage tiers is derived from preconceived hierarchies about how storage must be deployed in a given infrastructure. Given advances in software technology, storage virtualization, and advanced metadata indexing, tape is being adopted more often today as a capacity storage target for primary data. In an active archive, M&E environment, and many other large datacenters, tape is often relied upon heavily to reduce both footprint and energy consumption for primary storage. Active archive has boosted the use of tape in any datacenter, but a common misperception about it is that it must adhere to traditional HSM – Hierarchical Storage Management – classification restraints. This is absolutely not true. Either Tape can be presented as a file system for primary storage in active archives or LTFS in smaller environments. Under traditional definitions, primary storage is usually considered tier 1.
This evolution in tape’s use has made it no longer confined to just tier 3 storage or the final tier in an environment. In a modern datacenter, a single tape library might contain the primary copy of a file, the secondary copy of another file, and a tertiary copy of a third file. In this case, tape is housing 3 different levels of data within the same storage system. This brings to question…. what defines a “tier”? We often think of tape as tier 3 due to the fact that it was usually the “backend” of an infrastructure. However in an active archive, any storage platform can be a direct target for primary data that is still accessible to end users. In this case SSD, disk, and tape can all three simultaneously serve as tier 1 targets, depending on the retrieval time requirements of the data being stored.
When tape is labeled as tier 3 in an active archive, it asserts that a tier is not determined by the data path, but rather by the performance of a given storage system. This gets incredibly complicated in the world of modern storage equipment that often houses multiple media types within the same system. Also, given the flexibility of storage performance, it is very difficult to compare one system to another in a true “like for like” comparison model. Disk’s performance is inherently tied to capacity, but a single enterprise drive alone can only perform at 120MB/s. Disk systems are obviously capable of sending and retrieving data at much faster speeds than this, but drive to drive, tape’s 280MB/s with LTO-5 and 500MB/s performance with the TS1140 drive compressed, by far out matches any disk drive on the market today. Beyond this, the total number of drives, and other considerations determine a storage system’s ultimate performance. We all know that it’s safe to say the performance of tape vs. disk will vary in specific scenarios. Specifications are dependent on the variances of a given configuration and should not be used to determine the hierarchy, so it is unfair to assert tape is always tier 3. At the end of the day, how you design your datacenter will determine the hierarchy of the data path, but to generalize any storage platform as any particular tier is an arbitrary assertion--- not a fact. To permanently label tape as tier 3 is therefore also illogical, and does not belong in discussions about storage systems.
