« Glacier not as Death Nail for Tape »Said Spectra Logic.
By Jean-Jacques Maleval on 2012.09.05
Following the Amazon Glacier announcement, here is a comment by Molly Rector, EVP Product Management and WW Marketing, Spectra Logic Corp.:
I absolutely do not see Glacier as a death nail for tape. I think Glacier is a very interesting new offering and will likely be very successful. That success will not be at the expense of the tape market. In fact, I don't really see it as a very large threat to tape. Certainly tape has a strong position in archive and Glacier is targeted at archive. However, with exponential data growth and keeping that data longer, there is the need for constant evolution in products that help solve this archive storage problem. Glacier is clearly designed to be a deep cold archive. Tape's largest growth market is active archive where data is indexed and used frequently.
Glacier will likely work well for smaller data sets and for deep archive of data that is not likely to be accessed. The retrieval times and costs will be too expensive for actively accessed data. Tape will remain the go-to technology for active archive where it is very strong today.
A few examples of these markets include:
- Media & Entertainment - This is very large images in data centers with no specific requirement for large internet pipes. Typically archive hundreds of terabytes or many petabytes. M&E customers utilize their archive data on a regular basis. Actively querying it and pulling clips regularly. It is not practical at all for them to have restrictions on when or how they use their data. It is also not practical to move their data into the cloud due to bandwidth constraints. They could not get all their data up to the Cloud for archive affordably and need to use their archive data much more often than the Amazon service is designed for.
- High Performance Computing - These folks have multi petabyte data sets that are being created very rapidly. The data that is kept needs to be actively indexed and when needs to retrieved needs fast retrieval time. Typically archive petabytes of data and are now beginning to get to exabytes. I ran a panel at the Super Computing conference last year with several national labs and an analyst. The consensus from all was it was much more practical and much more affordable to move the compute to the data than the data to the compute. i.e. Cloud is not practical for HPC data archive. The presentation we walked through is attached. Please note that slide 29 summarizes the top reasons why is not practical for HPC.
- There are several new archive markets for tape that are emerging with attributes very similar to Media & Entertainment.
In summary, Amazon Glacier is not practical for the active archive markets that tape has enjoyed rapidly growing footprint in. Bandwidth costs are too high, retrieval is too slow and frequency of data access does not fit the Glacier model.
Additionally, the economics of Glacier are not competitive with tape. Glacier as I understand is priced at $0.12/GB/year.
The comparative costs for a T-Finity Tape Library are:
- Cost / GB / Year = $.05
- Cost / GB / Month = $.004
- Cost / GB / Month Amortized over 5 years = $.0008
Note of caution on the 5-year amortization figure: It does not include the power, floor space, or personnel costs in years 2-5 so it's a bit understated. However, even with those numbers included, the total cost should still be well below $.01 / GB / month for the period.
The configuration this was based on includes everything (encryption software, media, drives, installation, warranty, etc). No hidden costs and pricing was based on our average street price for T-Finity.
- 10-frame TFIN
- 10,000 slots (capable of going to over 11,000 slots for the frame count)
- 10,000 LTO-5 media
- 12 LTO-5 drives
- Pro Encryption (to protect the archived data)
- PS installation
- Preventive maintenance
- 1-yr warranty
- Power cost = $.08 / kwH
- Utilization (duty cycle) rate = 80%
- Floor space cost = $100 / sq. ft.
- 1 storage admin (high end of pay scale + benefits, estimated from on-line salary survey)
Glacier also does not solve many of the ongoing objections to Cloud for storage. Security concerns, concerns about placing key corporate assets outside the company, deletion concerns, data growing much faster than networks are speeding up, cost of bandwidth and the fact that for large data recoveries a truck with tapes is still the fastest way to retrieve offsite data. It takes about 40 days to move 1 petabyte of data over an OC48 (and most Cloud users don't get the luxury of an OC48 for remote data archive). All of these other objections remain unchanged.
I do think with data growing so rapidly and the need to keep data for long periods of time, new options for data storage are necessary. Tape's answer is Active Archive. Glacier's answer is cold deep archive. There is definitely a place for both to be quite successful in helping users with the data management challenges they encounter today.