Consider Tape Storage for Your Power Systems Environment
Tape is the only storage medium that's capable of delivering the volume of storage and performance read back to stage data for Power Systems analytic processing.
By Shawn Brume10/30/2019
The IBM “Performance Optimized With Enhanced RISC” platform has powered devices that have changed the world for nearly 30 years. In the 1990s, the POWER* chipset was the platform that Apple used to build a completely new generation of desktop computers. Now, most of the supercomputers that are crunching artificial intelligence (AI) research are based on the POWER9* platform. The IBM Systems team has always created synergies that help our clients meet their business outcomes, both directly and indirectly.
This synergy extends to the use of the IBM RISC chipset in other products—products that reach deep into the fabric of business and consumerism. IBM integrated the POWER processor and Linux* real-time OS (RTOS) as the primary mechanism for tape operation in 2010, sealing a long-term partnership in digital data control, compute and digital storage. Tape has been the leader in digital data storage for over 60 years. That may sound funny in modern terms, but IBM tape has always been agile in approaching business solutions. That agility has resulted in the sixth generation of digital tape consumption. How did we get to six generations?
- Generation 1: The only digital storage
- Generation 2: Near-line transactional storage
- Generation 3: Recall andretention data behind HDD transaction
- Generation 4: Backup and recovery (BaR)
- Generation 5: Archival storage and secondary BaR
- Generation 6: Secondary storage reducing expenses, ensuring regulatory compliance
It’s strange to imagine that tape technology still carries weight in a progressively software-driven world. However, it makes perfect sense: In order for AI to be effective, it must have historic data to access. Keeping all of the data in memory is impossible, and retaining the data on NVMe is monetarily inverse to the total value of the data. Operations working with direct Flash-to-Tape (FLAPE) infrastructures are finding that more often than not, data goes to an unused state faster than it can be tiered to HDD. That doesn’t mean there isn’t an opportunity for HDD, but AI and big data are going to see greater shifts to FLAPE.
Data Powers AI
That brings us back to the power of Power*. As the IBM Power Systems* platform continues to be the basis for the most powerful supercomputers in the world, focusing on those capabilities will monetize the vast amounts of data stored on tape infrastructure.
AI continues to change everyday life. All of that change is driven from the data collected and stored. Much of the advancement that will come from this space is from Edge and Internet of Things (IoT) devices. Today, only 10% of all Edge/IoT data is collected. Most of the data is thrown away at the device because of bandwidth—or, more than likely, because there’s no place to store the data. This data amounts to about 1.5 ZettaBytes (ZB) at present, but by 2021, this will jump to 85 ZB. We can use more analytics to optimize the data, but ultimately, we need more data.
The Value of Tape
Processing the data is the mastery of the POWER processor. No other technology has the capability to be both general- and specific-use at the same performance level when combined with NVMe. The problem of data storage is still an issue unless tape is embraced.
Tape is the only storage medium that’s capable of delivering the volume of storage and performance read back to stage data for Power Systems analytic processing. At 20 TB per cartridge and 400 MBps of streamed data rate, IBM tape is the single most cost-effective method for storing hyperscale volumes of data. Tape is also easy to use, standardized and secure. Like most technologies that are misunderstood, the idea that tape is easy to use has critics who are only countered with real examples. Tape examples should start at the top with hyperscale companies.
Hyperscale companies—Google (Alphabet Inc.), Facebook, Microsoft, Amazon, Alibaba, Baidu, Yahoo, Apple—are competing in the realm of economics of scale. Bigger is driving innovation in monetization of business assets, leading to new business in infrastructure as a service. According to the latest breakout, deep archive storage accounts for 7% of Amazon AWS revenue. In a business that’s a race to the bottom, controlling expenses
Tape Storage is Critical to Economies of Scale
According to the Cisco Global Cloud Index, by 2021, 53% of compute and storage operated in data centers will be located in hyperscale cloud provider locations. Tape storage will be critical to the continued economics of scale. POWER processors have enabled IBM tape to continue to be agile in deployments and offer enhanced usability. As performance and interoperability continue to play key roles in business operations, the Power Systems platform and IBM tape will be key partners in creating solutions.
As social and business delves deeper into the understanding of our daily habits, retaining more data on performance priced solutions will lead to all data ending up on tape. Regardless of which solution you choose—LTO-8 at 12 TB per cartridge or TS1160 at 20 TB per cartridge—either will retain data at near-zero energy consumption. The media life lasts longer than HDDs, resulting in a very low carbon footprint. As data grows, this becomes a significant part of meeting the three primary business goals of growing revenue, reducing expense and staying in regulatorily compliance.
Don’t get behind the “POWER” curve; investigate the IBM tape solutions today. A true hybrid cloud model requires having similar solutions on-premises and off-premises, which means that long-term data in your cloud is going to end up on tape someday.
Shawn Brume, Sc.D. is the global hyperstorage offering manager for IBM’s Data Retention Infrastructures team.