If Tape Is Dying It Has The Same Disease As HDDs
Recently Storage Newsletter, on the back of a report issued by Santa Clara Consulting Group (SSCG), stated that “Tape Is Not Dead, But Dying…”1. This opinion is based on worldwide tape sales, which declined 9% according to SSCG and the fact that the number of firms who are successfully growing their revenues as a result of selling tape systems has shrunk to only one – Spectra Logic. On the surface of it, this seems a very logical and well-considered view. However, in my opinion this is only because the issue of “Big Data” (there are others but “Big Data” seems particularly popular these days) and the growing desire by companies, governments and individual users (particularly those who are heavily involved in social media) to retain information forever has not been fully explored in terms of disk’s (and I am going to speak specifically to HDD) viability in being able to meet these requirements. While I do not intend to undertake an exhaustive look at this topic in this blog let me give you some food for thought as to why tape has the potential at least to make resurgence in coming years while HDD’s days are likely numbered.
The “Big Data” issue
EMC has spent considerable time promoting the phrase “Big Data”, tweaking the term to meet their particular marketing needs. Essentially, “Big Data” exists any time you get so much data, especially in a single data set, that it becomes difficult to manage. Other vendors are jumping on this bandwagon too hoping to capture the ever-increasing amount of money being spent on making “Big Data” immediately accessible and usable. But what about storing large data sets for a prolonged period of time? When “Big Data” consists of petabytes, exabytes or even greater amounts of data, is disk really the right solution? If we were only talking about keeping “Big Data” around for a short period of time then maybe but that isn’t the world we live in where, in a lot of cases, we would like to keep everything forever (Facebook anyone?). So, we face a huge amount of data that grows exponentially with a need to keep it for a long time. Now does disk make sense? If you answered yes, I suspect you either work for a disk vendor or enjoy spending copious amounts of money and keeping power companies in business through outrageous electricity bills … I invite you to read on anyway … No, disk doesn’t make sense when you start talking about keeping data around for a long period of time. That is the world that tape rules and for good reason. Just look at the typical costs for long-term data retention:
Long term data retention costs2
- Tape: As low as $0.02/GB/year; typically around $0.05/GB/year
- Disk: Enterprise-class SATA costs around $0.80/GB/year
These numbers show that even though disk has decreased in cost significantly it is still nearly 16x more expensive than tape for long term archival. Density doesn’t really play into disk’s favor either because while disk density has increased on average 36% since 20003 seek times have not improved more than 5-10% forcing drives to be used in arrays to achieve equivalent performance improvements driving cost per usable gigabyte up significantly. Tape, during the same period has, increased in density 15x and performance has improved 7x. (LTO-1 to LTO-5 at 100GB@20MB/sec and 1.5TB@140MB/sec respectively)4 all without experiencing the same type of usable space problems.
Infrastructure requirements
Disk also faces the challenges of requiring constant power and cooling which creates a scalability issue. After all, at some point a disk-only solution would consume all currently available floor space requiring a new facility in which to house additional disk systems to be built. That nthdisk system, then, becomes extremely costly and achieving a timely and acceptable ROI can be difficult at best (this is a problem that the “Cloud” faces today even though there seems to be very little talk on this point at present). Tape, by comparison can be stored pretty much anywhere (subject to controlling temperature and humidity, of course) making the facilities to house tape media far simpler and less costly to both build and maintain.
Portability
And then there is the matter of portability… Sure, you can remove a disk from an array, attempt to stick it in a box and move it to an offsite facility as a means of storing data for the long term but with the myriad moving parts inside of an HDD that probably wouldn’t be the best idea. Again, tape does not have this issue. Tape is designed to be portable and this, combined with the facility costs mentioned above, means data can be retained for a long time at relatively low costs and with relatively little risk to the data (assuming it has been handled properly in the first place). Furthermore, data on tape can be moved easily from one facility to another should that requirement arise and it doesn’t require a high-speed massive bandwidth connection to achieve that movement either, which is yet another benefit of tape over disk (and a whole other topic…watch this space).
Reliability
At this point you may be thinking that I am overlooking the issue of reliability because, after all, people will regularly cite tape’s failure rate at 70% or higher5. This is largely FUD. Sure, if tape isn’t handled properly then it fails but that is a quality you can ascribe to pretty much anything. For comparison, consider these statistics:
Error rates for various storage mediums6
- LTO Tape: 1 bit in 1 x 10E17 bits
- Enterprise Tape: 1 bit in 1 x 10E19 bits
- Desktop SATA: 1 sector in: 1 x 10E14 bits
- Enterprise SATA: 1 sector in: 1 x 10E15 bits
- Enterprise FC/SAS: 1 sector in 1 x10E16 bits
My experiences echo these numbers. Our IT team works with various disk arrays as part of several VMware ESXi deployments and they are regularly replacing hard drives in disk arrays and suffering performance degradations during rebuilds. Contrast that with our experiences using tape where, when proper care has been taken to store data on tape, ensuring error checking was done in the process (which most applications offer), problems with data restoration after the fact are few and far between. Where errors have been detected, which has been seldom, recreating a tape has been easy and the performance impact on other IT systems has been negligible.
In light of all of this, is it really fair to say that tape is dying? Well, maybe because the traditional use of tape is changing and tape drive and media sales are down. But, tape still has a critical role to play in the world of long-term data archival and in that respect it is healthy and has a potentially bright future as the desire to keep everything forever grows (which may even reserve the decline in tape drive and, to a lesser extend media, sales).
If tape is dying it has the same disease as HDD
Given the information presented herein I hope that you can see how tape has the opportunity to not only be a viable medium for years to come but, critically, to address the need for the long term archival of “Big Data” in a cost effective fashion. This is not, however, to say that tape is not at risk of being killed off it is merely to say that the threat, in this author’s opinion isn’t coming from HDD but rather from SSD; that tape and HDD are both fighting the same ailment which may claim the life of both technologies in due course.
Why is SSD the REAL threat?
SSD is a threat to both HDD and tape because it combines the best attributes of both technologies into a single package. SSDs are portable, capable of handling diverse environmental conditions (with the exception of extreme temperatures), can be stored on shelves rather than in arrays, provide extremely high I/O rates, etc. For these vary reasons SSDs are gaining popularity and in a study commissioned by IBM7 47% of respondents were either using or planning to use SSDs in the future. The one thing that is holding SSDs back right now from killing off HDDs and tape is cost. At around $2/GB for mid-range drives (enterprise class being even more expensive)8, SSDs leave organizations with the prospect of spending 2.5x times more than if they used HDDs and 40x more than if they were to use tape for the equivalent amount of data storage. With the cost of HDDs and SSDs likely to reach parity far sooner than tape, HDDs should be worried about their longevity. Tape, on the other hand, can breathe a bit easier … at least for now … as it continues to be the most sensible long-term data archival medium.
1http://www.storagenewsletter.com/news/tapes/tape-is-not-dead-but-dying
2http://www.linkedin.com/groups/is-TAPE-still-being-used-75803.S.54991695?view=&gid=75803&type=member&item=54991695&trk=EML_anet_di_pst_ttle
3http://www.infostor.com/disk-arrays/raid/2010/understanding-the-hidden-costs-of-storage.html
4http://en.wikipedia.org/wiki/Linear_Tape-Open#Generations
5Compuquip White Paper
6http://www.activearchive.com/content/archive-reliability-tape-vs-disk
7http://www.storagenewsletter.com/news/marketreport/ibm-survey-ssd
8http://www.anandtech.com/show/4421/the-2011-midrange-ssd-roundup
September 20th, 2011 by cyoung
Comments (1)
It is conceivable that the death of HDD to SSD will play out in the evolution and adoption of cloud resources. With further utilization of cloud resources and increased allocation of dynamic company data the requirement for point in time backup for recovery will grow into a quality of service problem for the cloud provider.
A problem that can be solved by combining all three technologies.