WD OptiNAND Technology for Hybrid Drive With HDD and NAND

Comments

Comments of Trendfocus:

This article was published on August 31, 2021 by Trendfocus, Inc.

Western Digital Launches OptiNAND Platform to Drive Capacity Increases
HDD

iNAND embedded flash drive (EFD), firmware and customized Soc reliably enable
higher HDD tracks-per-inch (TPI) without sacrificing drive performance

• Company states that OptiNAND is not a hybrid HDD as EFD is not used to actively store user data
• Repeatable run out (RRO) and write operations metadata to manage adjacent track
interference (ATI) are stored in iNAND rather than on disk and in DRAM
• Technology enables sector-level ATI refresh, reducing excess refreshes and enabling higher density track writes without additional performance penalty
• OptiNAND-enabled HDDs will also allow larger SMR (shingle magnetic recording) capacity gains over native CMR capacities

As the only company that currently maintains both in-house NAND and HDD businesses under one roof, Western Digital announced that future generations of its nearline, or capacity enterprise HDDs will incorporate a new flash-enabled architecture. Dubbed OptiNAND, the company is utilizing a BGA EFD (ball grid array, embedded flash drive) utilizing a UFS (universal flash storage) interface to manage HDD metadata in a more efficient manner than in current designs. In a pre-briefing last week, the company made clear that it was not deploying a hybrid HDD solution that has been used in other older generation HDDs to improve apparent write performance by storing incoming user data directly onto flash. The OptiNAND architecture in its first iteration is a 64GB iNAND package combined with customized firmware algorithms tied to changes in the drive Soc to move HDD metadata onto NAND.

In current HDDs, repeatable run out (RRO) information is stored on disk during factory configuration, while ongoing write operations to manage adjacent track interference (ATI) are handled by the DRAM in the drive. By moving both metadata types to NAND, some on-disk capacity is freed up (for RRO), while in the case of ATI management, the large capacity of NAND compared to DRAM allows the HDD to improve the intelligence of ATI refreshes. As described in the company's technical brief and accompanying blog post, as written tracks are packed closer together to increase areal density, information about those writes must be stored so that the data can be periodically refreshed. New data recorded on neighboring tracks will eventually interfere with previously written data, forcing these refreshes. In current HDDs, these data are stored in DRAM and refreshes must occur at the track level. Also, given the limited availability of DRAM capacity, only so much ATI data can be stored, so the drive will sometimes perform ATI refreshes on a best-guess basis. With the embedded iNAND possessing considerably more capacity than DRAM, much larger amounts of ATI-related metadata can be stored and thus ATI refreshes can be performed on a sector level, rather than at the track level. In addition, the rewrites can be performed much more intelligently. The net effect of this architecture is to enable even higher TPI, which will increase areal density for both CMR and SMR HDDs without causing significant performance penalties due to excessive background housekeeping tasks.

OptiNAND also increases the drive capabilities under emergency power off (EPO) events. The architecture can flush more than 100MB of write cache data during EPO, compared to 2MB in current drives. In write cache enabled (WCE) mode, OptiNAND reduces the frequency of write cache flushes, while random write performance in write cache disabled (WCD) mode can approach that ofWCE mode.

The first generation of OptiNAND HDDs will be a 20TB CMR nearline drive at 2.2TB/disk on nine disks, which is currently sampling to select customers. The helium-sealed HDD will also employ a triple stage actuator to assist in higher track densities. OptiNAND-enabled drives should lift the SMR capacity increase from 10% to somewhere between 15% and 20%, according to the company. While no formal announcement of an SMR variant of the 20TB CMR model was mentioned, Trendfocus expects the SMR capacity to hit 24TB based on market intelligence information.

Comments and Analysis:
Western Digital has also not publicly disclosed its long-term direction for energy-assisted magnetic recording (EAMR) technologies, only having mentioned in the past that it remains actively engaged in both microwaveassisted magnetic recording (MAMR) and heat-assisted magnetic recording (HAMR) development. The company currently utilizes what it calls ePMR in which a bias current is applied to the write pole, increasing head writing field. Used in the company's current 18TB CMR HDD shipping in high volume, ePMR will accompany OptiNAND in the 20TB CMR drive. Interestingly, Western Digital expects OptiNAND to extend ePMR densities well into future products, reaching a targeted 50TB in the second half of this decade. If this is truly achievable, then the urgency to transition to other EAMR technologies can be pushed far enough into the future to buy additional time to address the considerable technical challenges of commercializing either MAMR or HAMR.

Capacity increases for nearline HDDs are coming byway of increased component counts as well as additional costs to achieve higher areal densities. While the company did not disclose the cost of implementing iNAND, an estimate of the flash and controller costs along with the incremental adders of implementing triple stage actuators across 9 disks would be on the same order as the cost to add a 10th disk and 2 heads; however, a 10-disk OptiNAND product at its initial areal density would yield a 22TB CMR HDD. Western Digital is counting on improvements to ePMR coming from head and media optimizations which will combine with OptiNAND's ability to boost TPI to drive HDD capacities higher over at least the next few product generations.

All HDD manufacturers face increased HDD costs due to technology implementations and increased component content, along with the added depreciation costs associated with the capital investments to increase component manufacturing volumes. Supporting the nearline exabyte growth and long-term $/TB reductions demanded by hyperscale companies also means that HDD companies will need to achieve adequate product gross margins to invest in this growth. Rational pricing and ever closer coordination with cloud companies will become even more important in the coming years to sustain product supply and to realize the long-term value proposition of HDDs to provide cost-effective storage solutions for massive data storage requirements.

Comments of StorageNewsletter.com:

Eighteen hybrid drives incorporating 2.5-inch or 3.5-inch HDDs and flash chips were released between 2006 and 2015 by Seagate, Samsung, Western Digital and Toshiba with total capacities between 40GB an 1TB and flash between 128MB and 120GB. But there was finally no market for these units that disappear.

Here WD does not speak about "hybrid drives", probably not to mention these failures in the past, but there is some analogy with the new OptiNAND devices.

Today, the last lucrative market for HDDs is concentrated on nearline 3.5-inch units of the highest capacities. But, in this field, what the big users needing huge capacities want is essentially the best price per gigabyte. Of course, adding some flash chips will slightly increase the capacity but on a memory much more expensive, ending with a higher price per gigabyte even with relatively better performances. Acces time will be the same as regular HDDs as well as transfer rate limited by 6Gb SATA and 12Gb SAS interfaces, even if PCIe/NVMe is coming for HDDs.

Nevertheless David Hall, R&D engineering and veteran, WD wrote: "In some cases, you can get up to an 80% gain in performance. So now those customers that run write-cache enabled that were running up against I/O limits, if they get a lot more performance, they can use a lot more of that capacity."

He also explains: "To increase capacity, narrowing the write tracks too many times would cause an unacceptable performance impact. And that's where the NAND enhancement in OptiNAND plays a key role (...) By enabling write-cache, the drive can be a little more dynamic with its operations, because it can schedule writes and reads (and refreshes) in the most efficient order possible. Thus, performance will be faster. But there is that slim chance that, in a sudden power loss, the drive never actually has the chance to write the data to disk (...) with OptiNAND, the data will be saved to non-volatile NAND memory to prevent data loss."

The new drives will need a new controller to manage the exchange between magnetic and NAND memories.

Note that there are already flash in all the HDD controllers but with limited capacities (64K).

WD is in the best position to design the new OptiNAND drives as it is manufacturing HDDs as well as flash chips with Kioxia that could be acquired). It's not the case for competitors Seagate and Toshiba.

Finally as in the past, we don't see a bright future for hybrid drives even including OptiNAND, but, at the end, the market's reception will be the final judge.

WD is trying to expand its business comprising:

• 2.5-inch units between 250GB and 2TB but with no enhancements in internal devices since as far as 2018, no one apparently coming, and probably progressively disappearing in favor of SSDs. 2.5" enterprise drives were discontinued.
• 3.5-inch units between 500GB and 20TB culminating at 2.2TB per platter with 9 disks, with the expectation to reach 50TB in the second half of the decade. But these last years their capacity increased slightly and diminishing in percentage (from 16TB et 18TB to 20TB). Maybe you remember that in the past capacity of HDD were doubling each year.
• SSDs from 120GB (Ultrastar DC SA210) and up to 15.26TB on NVMe devices (Ultrastar DC SN840) and the coming use of BiCS6 and ZNS/Zoned Storage which enables increased capacities, improved QoS, increased endurance and reduced cost.