WLCG
Tape market shrinks/fails
The market for tape shrinks or fails completely, due to loss of revenues, shrinking markets, etc.
Major rethink of value of data vs re-run of experiment, Need new data models; Drastic reduction in overall storage capacity.
We would lose our traditional archive system (low cost, low power, low bit-error level, high density, high latency) for long term storage.
- This would imply a complete re-assessment in the HEP community of the value of data (and long term cost) vs the re-run of an experiment or reproducing the data (e.g. re-run MC, re-derive physics data).
- Would need new models of data lifecycle, new assumptions on data replication and assessment of cost vs risk
- Significant reduction in overall storage capacity
The future data lake model is part of a mitigation for this scenario. Focussing data storage on few large sites while delivering content to remote compute facilities would partly mitigate this. In addition, if we had to replace tape by HDD or SSD, would only be able to afford a single archive copy (of very selected data sets); this would change the overall level of availability of some data and how we treat the value of archives.
We have already seen Oracle pull out of Enterprise tape. Today we only have IBM as the manufacture of tape drives - both Enterprise and LTO (although heads are made in separate single company). The tape drive market is in the hands of a single company. Tape media is in hands of only 2 comanies who are in a patent dispute in the US; leading today to unavailability of latest tape media versions. Long term, traditional tape users are moving to alternative solutions. We may be forced to move to a very different alternative archive storage, changing our underlying assumptions completely.