"Tier 1.5 arrays" is a colloquial term used to describe block-based storage systems that are positioned in between Tier 1 arrays (e.g. EMC Symmetrix, Hitachi USP V and IBM DS8000) and modular, low-cost Tier 2 arrays (e.g. Nexsan, HP MSA, EMC CLARiiON etc.).
Examples of Tier 1.5 arrays include solutions from 3PAR, Compellent, IBM’s XIV, certain NetApp solutions and offerings from other vendors. Practitioners should note, however that arrays from these vendors differ significantly in scalability, performance, and function and serve different segments of the market. For example, 3PAR represents the gold standard of Tier 1.5 systems in more demanding workloads and competes with both Tier 1 and Tier 2 arrays. Compellent is positioned at the lower end of the Tier 1.5 spectrum and sets the mark for simplicity and ease-of-use within small and medium sized enterprises that would typically use Tier 2 arrays.
Nonetheless, Tier 1.5 storage arrays represent a new breed of storage systems that universally are easier to manage than traditional arrays and often deliver availability and performance levels that rival Tier 1 systems.
Tier 1.5 arrays can be described generally as having the following attributes:
- Storage Virtualization is a fundamental and native feature of these systems whereby an abstraction layer is provided that separates the physical hardware from the logical presentation of that hardware. This capability is typically implemented in some type of log-structured file.
- Data is typically spread across disk drives in the system to remove bottlenecks and provide consistent and reasonable performance. Storage is allocated on write.
- Overall ease-of-use, specifically for operations such as provisioning and data management, is a fundamental design criterion enabled by the virtualization abstraction layer.
- The ability to dynamically move data within certain boundaries, without disrupting an application’s access to storage, is enabled by virtualization. As such, storage placement is non-contiguous, unlike in traditional LUN-based arrays.
- The introduction of an abstraction layer typically makes it harder to optimize performance on individual LUNs, but facilitates the ability to deliver "good enough" performance for all LUNs.
- Features and functions are very rich and take advantage of a virtualized infrastructure (e.g., virtual or space-efficient copies, snapshot-based replication). Usually these capabilities are greater than those found in Tier 2 arrays.
- Modular components - When tier 1.5 systems scale out, the typical design is scale-out NUMA. Rather than a single, fast and expensive shared memory (e.g. SMP), processors and memory are separate building blocks and access times to memory/processor resources are variable (i.e. non-uniform) across a fabric.
- Controller designs are loosely clustered when scale-out is required.
- Typically Tier 1.5 arrays have lower acquisition and operating costs (CAPEX and OPEX) than Tier 1 but higher CAPEX than Tier 2 arrays.
- Reliability, availability, and serviceability (RAS), and remote replication are typically less robust than Tier 1 arrays.
Tier 1.5 arrays do not typically compete with Tier 1 arrays for the very highest performance, very highest availability workloads, however Tier 1.5 systems are siphoning many workloads off Tier 1 arrays. On balance, Tier 1.5 arrays stress affordability and better price/performance than Tier 1 storage arrays and provide a simpler management schema. But the most complex workloads and demanding RPO and RTO requirements should still be allocated to Tier 1 arrays generally.
The bottom line on Tier 1.5 systems is they typically provide greater functionality and scalability than Tier 2 systems and have become a viable alternative for small and mid-sized companies that have not been able to cost justify Tier 1 storage in the past.
Action Item: Tier 1.5 arrays are here to stay. Their 'good enough' performance attributes are suitable for the majority of block-based workloads and place these systems in the sweet spot of value-added storage arrays. Customers should evaluate the benefits of simplified management and solid recovery as compared to lower cost, less functional Tier 2 arrays and much more expensive and mega-functional Tier 1 arrays, to determine their proper strategic fit.
Footnotes:
