One of the most daunting tasks facing the HPC industry today is how to manage the onslaught of exponential data growth rates. As users create, gather and move various data assets though their life-cycle, multiple applications are used to access, process, manipulate and output the data.
It is critical that the storage infrastructure is configured and tiered appropriately to enable an efficient workflow that allows timely and cost-effective access to data. New tape technologies combined with disk, and eventually flash, can address the current and future challenges and objectives of the HPC storage administrators, including performance, compliance, security, cost, energy consumption, asset preservation and data protection.
The introduction of a Linear Tape File System (LTFS) has paved the way for tape network-attached storage (NAS), because it provides the ability to access files on LTO tape in the same way if they were on disk. Let's take a look at some of the reasons to implement tape NAS.
On average, the cost differential between disk and tape is holding steady at around 10 to 1. For many applications this is acceptable. This differential would apply to the data that should reside on Tier 1 storage because of its critical nature or high speed accessibility requirement. The value and access to the data is equivalent with the cost to store and manage it. I find that this data typically represents about 20% of most customers' data volume.
However, one must consider the entire cost of Tier 1 storage including the capital expenditure to acquire it, along with the operational cost to support power and cool it, and the space to house it. Best practice is to make multiple copies of Tier 1 data regularly and incrementally as it changes and manage it through intelligent, policy-based file system applications.
As we move down the pyramid of storage, we come to what I call Tier 2 of storage. This is where tape is gaining momentum and encroaching on what used to be primarily disk territory. With today's tape technology attributes, more and more data managers are finding tape to be the answer and solution to their biggest challenges regarding the storage tsunami. Tape has a long media life, higher reliability than disk, and a significantly lower cost. This has made it optimal for offline storage and data protection and archive.
Historically, tape has dominated in this role and has been the technology of choice. In recent years, however, because of the continued innovation of tape from various vendors, it is now being positioned as a Tier 2 NAS. These innovations include:
- The ability of tape to outperform disk when architected properly.
- The ability of tapes to be written to and read from through a native files system in file format transparent to the end user, aka NAS/LTFS.
- The greener impact of tape storage and increasing power and space costs.
- Total cost of ownership improves storage economics.
- The significant capacity increases in tape, which makes the cost per GB or TB 10x less than disk.
- The reliability and lower bit error rate (BER) of tape over disk.
Aside from the 10x lower cost of tape and the increasing importance of its lower bit error rate, tape also has a useful life of at least twice as long as disk drives.
LTFS has become a game changer in the world of storage and will allow users to leverage the full capability of today s tape. The LTFS specification was created by IBM and is available as an open standard for standalone tape drives using LTO.
Basically, LTFS allows tape to be used in a fashion like disk or other removable media making tape the most cost-effective archival data storage solution for most applications. Continued development and innovation in tape indicates that this trend will continue well into the future.
I'll be exploring this topic in more detail at the SGI User Conference in San Diego on May 2nd. If you are in the area, come check it out!
