Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
Some storage vendors have been predicting the death of tape since the mid-1990s, when 3.5-inch disk drive densities started to increase.
I remember a sales manager for a storage vendor telling me at the time that a customer was "just plain stupid" to have its 30 TB archive on tape and should be buying spinning disk instead. The customer "just doesn't get it — tape is dying," the sales manager said.
Well, it is more than 10 years later, and this same person is now at a different company selling tape solutions, and the customer now has a 2PB archive on tape.
No matter how many silver bullets or wooden stakes disk storage vendors try, they haven't been able to make tape go away. It is still here, and I think it is going to continue to be here for the foreseeable future, high-profile tape losses notwithstanding. Issues such as cost, capacity, power, portability and bandwidth will see to it.
Here is a chart showing Seagate Disk SCSI/Fibre Channel capacity growth over the last 15 years:
It looks very impressive until you look at the capacity growth rate on an annual basis:
As you can see, there has been a sharp deceleration in the capacity growth rate over the last five years, and it's only going to get worse. The next disk density jump is likely to be to 400 GB Fibre Channel drives, not 600 GB Fibre Channel drives, so even the 15-year trend of doubling densities is likely to change. The point, as we shall soon see, is that disk's capacity advantages over tape may be starting to erode.
Power is another issue. In looking at the Seagate Web site, a Fibre Channel 300GB drive typically consumes about 18 watts of power while in use; idle power usage is about 12.2 watts. We will talk more about power issues later.
The density increases in tape have not matched those of disk over the last 30 years, looking at uncompressed capacity. Compression chipsets have gotten bigger and faster, but as all of the vendors say about compression, your mileage may vary. Some data types such are not compressible, so I think it is important to examine raw capacity and look at compression as icing on the cake.
Here's table showing the growth in tape capacity over the years. If multiple drives where introduced in the same year, the highest capacity drive is listed.
Tape capacity has grown in fits and starts, but over time, it has kept up pretty well. Imation and the LTO consortium have announced plans for higher density tape technology, but roadmaps are roadmaps, and until roadmaps become technology you can buy, don't count on it. Improvements in density technology do not have a good track record for being on time.
It is important to note that tape has no power requirements, unlike disk drives.
Power, Reliability and Portability
Let's take an average site that has 100TB of disk storage, and let's say for purposes of efficiency that this site uses 4+1 RAID-5 for the LUNs and 300 GB drives. The power to run all of the disk drives can be calculated by:
100 TB of storage/1.2 TB per LUN (4 times 300) = 84 LUNs, rounded up.
That's 220 drives (84 times 5), and you should add in some hot spares, so say we have 225 drives. Using the Seagate data, that would be 4.05 kilowatts (18 watts times 225 drives), not including the RAID controller, the air conditioning and fans to cool the heat. The amount of power for this is often far greater than the power for just the disk drives, depending on time of year, location, humidity and many other factors. Given the rising cost of energy, this is no small consideration.
For tape, you have no continuous power costs other than running the tape drives and the robot. This amount of power is virtually nothing compared to the power needed to run the disk drives.
There are many who say that tape is not as reliable as disk, but reasonable arguments can be made on both sides. The question I always ask is how many people ship disk drives through the mail? Tapes are mailed all the time, and the portability is part of the media's attraction, but they can also be more easily stolen or read. Disk drives have far more mechanical parts, and thus can suffer from too much shaking, rattling and rolling.
Tapes can be ingested into robots easily from shipping containers specifically made for tapes. Shipping disk drives and having a backup application that can use them is a far greater challenge. This might change in the future, but tape has a big portability edge today.
What will the future hold? New technologies such as MAID (massive array of inactive disks) are starting to be sold by a number of companies, but MAID does not offer the ability to ship data around the country via the carrier of your choice.
The cost of network bandwidth prohibits most organizations from moving data around the country and the world at the rate that it is created. Network performance over WAN connections are not growing as fast as disk and tape density.
I believe that tape will not be totally replaced by spinning storage technology, given the cost of power and WAN connections. Yes, MAID might take the place of first copy in some environments, but no one can ship 100 disk drives and figure out what the mount and mkfs parameters are for each drive, and there are no carriers for allowing disk to be shipped like tape anyway.
Libraries and backup/restore applications know how to manage each tape device, and tape devices have better bit error rates than SATA drives and even Fibre Channel devices. No one knows for sure what the future holds for any technology, but I am pretty sure that tapes will be in our technology future for as far ahead as I can see.
Henry Newman, a regular Enterprise Storage Forum contributor, is an industry consultant with 24 years experience in high-performance computing and storage.
For more storage features, visit Enterprise Storage Forum Special Reports