By Robeel Haq
Cinegy's Igor Petrov looks at the arguments for and against tape and disk as a storage medium.
Cinegy's Igor Petrov looks at the arguments for and against tape and disk as a storage medium.
At the beginning of the 21st century, hundreds of articles were written on the topic of ‘tapes versus disks'. The followers of the two opposing parties have been trying to convince us with zeal of the advantages of this or that technology, weighing in with various ‘for' and ‘against' arguments.
Time has passed and the debate has quietened down; in the TV industry, however, the controversy continues. Today, in the aftermath of the financial crisis and a reduction in the volume of content produced, people have taken a keen interest in digital archiving technologies, raising the question: What is the best way to store content?
To come up with an answer, it is necessary to separate the reality from the mass of marketing myths that influence our subconscious with frightening stories about ‘hidden expenses'. I have talked with many advocates of tape libraries, and their arguments in favour of tapes are varied, but the major points are ultimately the same: the first being the lower cost of tape storage, the second being the higher reliability of a tape cartridge.
Let's take a good look at each argument to see what foundation these arguments have in reality. While considering this topic, I have intentionally not named specific models and their specifications, in order not to advertise one or offend others. It is sufficient to say I compared product prices and technical parameters of such companies as HP, IBM, Quantum, Overland, Arena (Maxtronic), Infortrend (EonStor), Xyratex, NexSan, Western Digital, Seagate, Fujifilm.
So let's start: The cost of a storage system is the sum of the following costs: the medium; reading and writing devices; power consumption and maintenance.
The maximum available volume of a tape cartridge today is 800 gigabytes (LTO-4 standard) at a price of around $70 each. Odern hard disks, SATA-II/SAS, have a maximum volume of 2 terabytes and are available for $200-$350 each, depending on their reliability level. More expensive models have a larger safety margin and a special code: RE (RAID Edition). With a few easy calculations, we get a twofold difference in price: 8 cents for 1 gigabyte on tape (LTO-4) and 10-18 cents for 1 gigabyte on hard disk.
Reading and writing devices
Storage system with 40-45 terabytes:
A disk array with an external FC-4GB interface and 24 2-terabyte disks, SATA-II RAID Edition (45 terabytes in RAID-6 altogether) costs around $24,000.
A tape library with an FC-4GB interface for 60 slots, one LTO-4 drive and 60 cartridges (48 terabytes altogether) costs around $25,000.
We get roughly the same number; however, this is just on the face of it. There is an important point, which the producers of tape libraries often forget, namely that a tape library does not work without the software to manage it (which in the case of a disk array is not needed).
This changes the situation significantly, as the software license for tape library management is not much cheaper than the complete tape library package itself, and sometimes is even more expensive.
As a result, a 45-terabyte disk array (DV25 3,500 hours) as a complete package costs half the price of a tape library of the same volume.
Storage system with 80-90 terabytes:
A disk array with an external FC-4GB interface and 48 2-terabyte disks, SATA-II RAID Edition (90 terabytes in RAID-6 altogether) costs around $50,000.
A tape library with 100 slots and two LTO-4 drives and 100 cartridges (80 terabytes altogether) costs around $60,000. Once again, the managing software doubles the library cost - so the result is at least $120,000.
In spite of doubling the volume, a 100-terabyte disk array (DV25 7,000 hours) costs half the price of the tape library, as before.
Storage system with 180-200 terabytes:
No doubt, an easier way of calculating the cost of storing 200 terabytes on disk arrays is to count up several separate devices of 20-25 terabytes in volume. However, we started comparing separate devices so we'll continue in the same way.
Already today, there are disk systems with 180 terabyte capacity in one box. In reality, it is the same 48-disk array with the add-on module for 48 disks more. As the cost of the add-on module is practically equal to the cost of the main module, we get a total cost for the system of no more than $100,000.
A full tape library package with the cartridges for 250 slots and with 4 LTO-4 drives will cost the same. However, as before, the cost of the managing software equals the cost of the library itself. All in all, we get about $200,000.
Once again, for a 200-terabyte disk array (DV25 15,000 hours) a double cost ratio in favour of disk arrays is preserved.
Storage systems of 500 or more terabytes:
Unfortunately (or fortunately perhaps), very few TV companies can boast of having 40,000 hours of ‘useful' content. For equality's sake though, let us define the threshold after which the tape library cost equals the disk array cost.
Calculations show that a distributed storage system on disks (SAN/NAS) "catches up with" the similar volume in a tape library after 600 terabytes, due to the fact that, fortunately, the tape library management software cost does not have linear dependence on the system volume. So with large storage volumes it makes up a little less than half the price of the hardware component.
As a result, with a volume of 1 petabyte (DV25 80,000 hours) the cost of a complete library package including managing software nearly equals the cost of an equivalent system with similar volume based on hard disks.
Until recently, the power consumption point was an argument entirely in favour of tape cartridges. However, we are now seeing a general adoption of green technologies that has not been overlooked by the storage system providers either.
A ‘green' 2-terabyte hard disk with the SATA-II interface and spindle speed of 7,200 revolutions a second uses about 3.5 watt in ‘idle' mode, about 7 watt when ‘reading/writing' and at least 1 watt in ‘standby' and ‘sleep' modes.
A 48-terabyte FC-4GB disk array (24 2-terabyte SATA-II disks) filled with such disks, as a rule, has two to three 350 watt power units for ‘hot' reservation. Tape libraries with 48 - 60 cartridges, as a rule, have one power unit that uses the same 350 - 400 watts.
As for the maintenance cost, here everything is more straightforward. Finding any system administrator with experience in working with tape libraries is much more complicated than finding a really good system administrator who knows all the peculiarities of working with disks and disk arrays.
Besides, tape libraries are usually sold with the obligatory technical support from the producer, the price of which as a rule, by far exceeds any evident or hidden costs associated with disks.
The reliability of a storage system is a combination of the following factors: system lifecycle; ease of data recovery and the impact of the external environment.
I have often heard the claim that "hard disks break down more often than tape cartridges", as a main argument in favour of tapes. In the course of the conversation, it turns out that we are talking about hard disks in the NLE stations and tape cartridges on the shelf in the archival room, and at the same time nobody considers that a hard disk in the NLE station performs thousands of read-write operations each day, and a cartridge just lies there and collects dust.
Referring to the objective information, we will see that the MTBF (mean time before failure) value of hard disks is 1.2 million hours, compared to that for the LTO-3 and LTO-4 cartridges where this length is only 250,000 hours, and the average life time of the tape drive magnetic head is even smaller - 60,000 hours.
This means that under the same maintenance conditions, hard disks are four to five times more reliable than tapes, and 20 times more reliable than tape'smagnetic heads. The fact that hard disks should be changed every 3-4 years in order to avoid their mass failure is indisputable. This point, of course, is the main argument in favour of tape's ‘durability'.
However, let's try to understand which durability we are talking about. Magnetic tape has been used in television since the beginning of 1950s - at first on reels and then as cassettes (since 1969). Disks appeared only in the 1970s and found their application in television just at the end of the last century.
Even within this short period, both tape and disk mediums have changed in terms of formats and interfaces. For example, LTO-1 tape cartridges were standardised only in 2000 and the LTO-4 generation only in 2007. Hard disks with the SATA interface appeared in 2003 and SATA II in 2005.
What durability are we talking about if we cannot even be sure that in 10 years there will be a device capable of reading our ‘durable' medium? In my opinion, the system durability today should be defined not by the moisture resistance and magnet security parameters of a separate medium, but by the system usability and simplicity from the point of view of migration to new storage generations and technologies. Disks, thanks to the speed of reading and simultaneous access, win again.
Ease of data recovery
Another argument in favour of tape: Data can be restored from tape even if the tape is broken or damaged, while it cannot be recovered from a damaged disk.
Expert opinion refutes this statement. A British company, specialising in restoring data from any damaged medium, says that the difference between disks and tapes is not that big, saying that over 50% of data from a tape (as well as from a disk) cannot be restored.
The external factors that could impact the condition of a disk or tape could be climate, dust, magnetic pickups and mechanical effects. Hard disks are not leak proof (despite a widespread belief that they are), but it is more difficult for dust or moisture to get inside than with a tape cartridge.
Both disk and tape use the same method for recording; therefore, strong electromagnetic current is equally destructive for both.
The only argument in favour of tape is obvious - a tape cartridge is lighter so it is more resistant to physical damage. In production processes where data movement is associated with medium movement from one place to another, this is very important. But is this relevant if a cartridge is simply archived deep inside a tape library?
We have reviewed the major arguments from tape advocates and it seems the majority of these arguments do not hold up in reality. With small storage volumes (up to 40,000 DV25 hours), a tape library costs twice as much as a disk array of a similar volume and it does not provide even an approximation of the speed and reliability offered by a disk array.
With larger storage volumes (from 40 to 100 thousand DV25 hours), a tape library costs as much as a disk system. At the same time, reliability of disk storage exceeds tape reliability and cost savings on power consumption when using tape will be far less than the annual salary of a tape library expert.
Sometimes a list of arguments in favour of tape can extend to claims that a tape library takes less space and that the reading/writing speed of tape cartridges is the same as for disks. These arguments are not supported.
A tape library on 60 slots takes up about 1 OU (libraries on 150-200 slots occupy the whole rack), and the height of the disk array does not exceed 4U. The maximum reading speed for a tape cartridge is 90 Mb/s (not taking into account time for rewinding and searching for a necessary segment), compared to any disk array today which ‘instantly' reads at the speed of 300 Mb/s.
Not only do hard disks provide immediate access, but also multi-user access to the content. This eliminates the inevitable waiting that occurs with tapes, where queues for the material make it almost impossible to either efficiently or conveniently use a large archive, such as in a large TV company. The real purpose of a TV archive is not just storing the content, but reusability of the accumulated material.
The result of the ‘tape versus disk' duel is clear today; we have not found any reasonable advantage in using tape libraries to store TV content. Television, slowly but surely, is becoming tapeless, and the same lot awaits digital archives in general. Taking into consideration the onrush of solid-state recording technology, we may well witness SSD technology taking over and rendering hard disks obsolete. At present, however, disks are the most reasonable choice when building up storage systems for TV content, whether of small, medium or large size.
Igor Petrov is the regional manager, Russia and CIS, at Cinegy GmbH. Visit: