It’s Time for Hard Drives to Join Tape In The Archive Tier

It’s Time for Hard Drives to Join Tape In The Archive Tier

Chris EvansCloud Storage, NAS Storage, Object Storage, Storage Performance

It’s Time for Hard Drives to Join Tape In The Archive Tier
Hitachi Hard Drive, picture taken at Hitachi uValue in 2010

This week we’ve seen the announcement of 14TB 3.5” hard drives from one vendor and 12TB from another. It’s hard to imagine the concept of 14TB in a single hard drive when ten years ago we were looking at only 1TB devices. However, these devices have a problem. The throughput and latency has remained relatively constant during the last 10 years. 7.2K RPM drives from ten years ago had a throughput of around 120MB/s at best (sequential). Today, we see around 220-230MB/s, but individual I/O latency for the newest drives (which use SMR) is about 50% worse for writes than they used to be.  It’s time hard drives were permanently relegated to the archive tier with tape.

The Performance Problem

Hard drives have been an amazing technology.  From the origins at IBM in the mid-1950’s, drive manufacturers seem to have been able to overcome seemingly insurmountable physical challenges.  As I discussed in a recent post, the industry hasn’t quite reached 100TB drives, but we’re moving there quickly.  However, we continue to scale capacity exponentially without the same increase in performance.

The market for hard drives has split in two.  Capacity drives continue to push on with capacity increases, albeit with worse latency.  At the high end, performance drives have effectively stalled in capacity because the market is moving to flash where SSD capacities far exceed hard drive models.  The only possible reason for retaining hard drives (in either category) is cost.

Cloud Storage

The divisions in the market are highlighted in the recent podcast we recorded at Storage Unpacked.  We talked with Andy Hardy, Regional Vice President at Nasuni.  I haven’t spoken with the company in some time and in fact haven’t written about them in perhaps 3-4 years.  I’ve posted some links at the end of this post, but as background, Nasuni is essentially a global NAS solution, deployed on-premises with the data stored in the public cloud.  Andy raised a number of interesting points during our podcast.

  • Dollars are with the IOPS – this is obvious from a tiering perspective, but even more accentuated with the deployment of flash.  I/O performance costs money, so optimise the data on fast media.
  • Businesses have a 80/20 rule for data – 80% is unstructured (object/file), 20% structured (block).  It’s likely this ratio will change in favour of more unstructured content.
  • The CFO doesn’t like unnecessary spending.  For global companies, ensuring IT is deployed in the right place can be hard and wastage exists.

Have a listen (and subscribe) to the podcast for the full story, but we can see Nasuni as ratifying the “Flash and Trash” message that Enrico Signoretti highlighted a few years ago.  Of course this isn’t suggesting that data on spnning media is trash, but has a different way to be treated.

The Architect’s View

Nasuni isn’t the only cloud storage vendor in town, but their experiences show that it is possible to get data out of the data centre and into Public Cloud.  Even with on-premises object storage, most data can be served out of cache, with HDDs providing no more than an active archive.  I wonder how much longer the HDD manufacturers will keep selling performance devices and how much of that business flash continues to take.  If I was doing business transformation today, I would be moving away from on-premises deployments as quickly as possible.  The only exception would be where data needs to be onsite for compliance.

One last thought; next month I will be at Storage Field Day 14 listening to INFINIDAT, a company that has built a modern storage array on hard drives.  InfiniBox bucks the trend the industry is taking for performance IOPS platforms.  How will they continue to grow as hard drives slow down?

Further Reading

Comments are always welcome; please read our Comments Policy.  If you have any related links of interest, please feel free to add them as a comment for consideration.  

Copyright (c) 2007-2020 – Post #3B6E – Brookend Ltd, first published on, do not reproduce without permission.