Excerpted by David Reinsel, Director, Storage Hardware Research,
IDC
Note: The following is an excerpt from an IDC publication¹
Hard disk drive (HDD) interface transitions within enterprise systems are being clouded by a number of issues. Disk storage system vendor strategies are far from similar and HDD vendors are considering various options with respect to interface technology transitions. One thing is certain: end users will not tolerate a digression from today’s performance and reliability. Moreover, they expect to see cost per gigabyte continue to drop.
As the parallel SCSI HDD interface finishes its longstanding run for over 20 years, a relatively new interface, Serial Attached SCSI (SAS), is prepared to take the baton. The difficulties of extending the legacy parallel SCSI interface are well-known, and SAS is being greeted with open arms. SAS drives have been available since 2004, but ‘real’ shipments started in 2005. Over 3.7 million SAS HDDs shipped worldwide from 2005 through 2006.
Small form factor (SFF) enterprise drives, originally positioned as 2.5inch 10,000rpm ‘performance’ drives, are now the likely successor to 3.5inch 10,000rpm in most scenarios. Price parity at any given capacity point has nearly been reached and in some cases, 2.5inch 10,000rpm drives have all but replaced 3.5inch 10,000rpm drives in various server product families.
Interestingly, all 2.5inch enterprise-class HDDs today are integrated with native SAS electronics. This poses no issue with server storage since the transition from SCSI to SAS is well under way. However, no such transition has started in external storage, where Fibre Channel (FC) drives are firmly rooted.
FC disk drives started to ship in earnest beginning in 1997, when unit shipments exceeded 100,000 units (not even 1% of the HDDs consumed in the enterprise). Today, FC drives represent around 21% of all drives shipped into the enterprise. In 2006, IDC estimates that over 7.5 million FC drives shipped to enterprise system customers.
While the HDD interface migration path is apparent with respect to internal storage, external storage is where the migration path is less clear. Overarching themes for external storage include higher capacities and lower costs. But that does not equate to the unilateral use of Serial ATA (SATA) drives. SATA drives have been proven acceptable in many enterprise storage solutions. The SATA interface, however, has not.
Deployed properly, enterprise-ready SATA drives provide an excellent solution to meet rapid storage growth in enterprise datacenters cost effectively. The key benefit from SATA drives is not with the SATA interface, but instead the higher storage capacity per platter that is possible compared to the capacity per platter in enterprise class drives. Given the price per GB advantage of SATA drives over enterprise class drives, their use in enterprise storage is nearly guaranteed.
Drive manufacturers often are faced with difficulties when introducing new technology. Customers rarely race to pay more for disk drives, even given their strategic place within an enterprise system architecture. Nevertheless, whether it’s integrating a new interface technology or form factor, HDD vendors must recoup the cost of this R&D investment. Typically, the HDD return on investment (ROI) reveals itself in terms of a price premium, hence slowing customer adoption.
Drive OEMs must manage an increasingly diverse set of customers that possess multiple needs and strategies with respect to enterprise system designs. The location of enterprise storage (internal or external to a server or in a storage area network [SAN]) can determine, or at least influence significantly, the type of HDD adopted. Therefore, not all suppliers have the same motivation or decision matrix when considering next-generation HDD technology. Consequently, HDD companies must manage a set of HDD features that include form factor, interface, spindle speed, and capacity.
Given the plethora of possible form factor and interface options, the question is: What will HDD interface and form-factor mixes look like over the next few years? Long-term, we think 2.5inch enterprise drives will become the lion’s share of drives spinning at 10,000rpm and faster. What interface is used in the future is less clear, although IDC believes SAS has the advantage. At the same time, displacing 3.5inch drives with 2.5inch drives in high-capacity storage arrays will be tough because of the phenomenal capacity and cost per gigabyte advantage of 3.5inch HDDs over 2.5inch HDDs – at least for now.
A 2.5inch enterprise hard drive consumes approximately half the power of an equivalent capacity 3.5inch drive. However, this does not guarantee a 2.5inch-based storage system will be more energy efficient than a 3.5inch-based system. There are a number of variables to consider in order to make that determination.
Another area of confusion for the industry is the product roadmaps for 3.5inch enterprise drives. With the looming potential of a choice of SAS HDD form factors and spinspeeds being offered by the HDD industry, the pivotal question now becomes: Which RPM for 3.5inch HDDs will system OEMs want? And, will more than one HDD OEM (either existing enterprise HDD vendors or those that may enter the enterprise HDD market segment) offer high capacity 3.5inch 10,000rpm drives with a SAS interface? As long as the potential sustenance of 3.5inch 10,000rpm SAS drives remains real, we believe this will be a negative influence to the adoption of 2.5inch SAS drives by some (possibly major) system OEMs.
Generally, SATA is synonymous with capacity. However, before SATA was fully tested and proven reliable for the enterprise, certain companies desired a high-capacity, low-cost FC HDD. This was achieved a couple of different ways, but included the use of native FC electronics on the HDD. In doing this, a system OEM could leverage the advanced features of the FC protocol (e.g., dual port, full duplex, better error handling, and better signaling) and high-capacity, without the cost and potential performance burden of bridging SATA to FC. The issue with this strategy was that there were only two suppliers of this ‘FATA’ drive, and then one dropped out.
The low-cost, high-capacity FC solution today is achieved by bridging a native SATA HDD to FC by using an additional component for each SATA drive. Many external storage OEMs have gone this route very successfully. The advantage of this scenario is that a system OEM can leverage the large base of SATA HDD OEMs and is not tied to just one supplier. This can be quite advantageous when certain HDD OEMs reach higher capacity points more quickly than others. It also mitigates exposure to a potential single-source.
With this in mind, is it possible that system OEMs will want to do the same thing with SAS? Even though SAS can control SATA drives natively, there are still weaknesses with the SATA interface in an enterprise computing environment (again, the lack of dual-port, full duplex, signal integrity, etc.). The question is, would the industry (HDD or system OEM) prefer to bridge SAS-to-SATA for each drive? Or, would the preference be for a high-capacity, low-cost SAS drive using native SAS electronics on a high-capacity SATA-like platform? Another decision pending within the enterprise HDD segment.
In the end, IT and datacenter managers have grown accustomed to increasing enterprise system performance and reliability in conjunction with a decreasing cost per gigabyte. System OEMs and HDD vendors will need to work together closely to continue along these same metrics. IDC believes that a single approach to SAS interface configurations would best serve the industry, and ultimately end-users. We expect to see the current murkiness related to enterprise HDD interface transitions to linger throughout the year, before a more crystallized path, favoring SAS and 2.5inch HDDs, emerges in 2008.
Figure 1: WW HDD Shipments into Enterprise Solutions, 2005-2010
Note: Figure 1 represents IDC’s current thinking regarding the industry’s adoption path of HDDs consumed within enterprise systems.
¹ IDC #205612, ” HDD Interfaces in the Enterprise: Sorting through the Murkiness”
