By Jane Clabby, Clabby Analytics
When we think of data storage today, chances are that the first things that come to mind are “object storage”, “cloud storage”, “commodity storage” and “software-defined storage”. Nobody really talks much about SANs and block storage anymore–except for high-end database and transaction oriented applications. The same has been true for IBM’s storage strategy.
For the past couple of years, IBM has made many announcements and published many marketing documents that position its Storwize and Spectrum portfolios, including “all-flash” versions of those products. However, the DS8000 family is typically relegated to the background, and enhancements have focused primarily on “speeds and feeds” – x times faster, x more capacity, x times lower latency etc. – rather than new use cases and support for different types of workloads.
Imagine my surprise when IBM’s first storage briefing of 2017 highlighted not only the new all-flash DS8880F arrays, but also the strategic nature of the DS8880F family and how it underpins IBM’s cognitive computing and analytics offerings. In addition, this announcement provides a lower-cost entry point (starting at 95,000 USD) for DS8880-class storage, providing high-performance low-latency storage for mid-range customers. According to IBM, providing a more cost-effective offering enables businesses in emerging geographies such as Latin America and Eastern Europe to afford DS8880F benefits in a smaller package.
New use cases/workloads
Here are a couple examples of workloads that are ideal for the DS8880 family:
· Cognitive computing
IDC forecasts global spending on cognitive systems will reach nearly $31.3 billion in 2019 with a five year compound annual growth rate (CAGR) of 55%. With IBM Watson, IBM has been a pioneer in cognitive computing –using Watson-based artificial intelligence technology in applications across a wide range of industries, enabling humans to solve complex problems more efficiently. Cognitive applications ingest, process and correlate huge volumes of data which requires a robust computing infrastructure and high performance, high capacity storage.
· Real-time analytics
Another area where IBM is seeing interest from customers is in real-time analytics, where information can be collected, processed, and analyzed in real-time for applications such as credit card fraud detection and Internet of Things (IoT). The DS8880F provides the robust resiliency and data security for the source OLTP data set as well as the throughput and performance to enable real-time analytics. As a result, financial institutions can match credit card use patterns with real-time data collection, detecting potential fraud before it happens and saving millions of dollars. In transportation, IoT data collected from cars, street lights, cell phones and other sources combined with real-time analytics will enable driverless cars. In health care, patients can be monitored–collecting and analyzing several metrics simultaneously in real-time– to alert medical staff to potential problems, improving patient outcomes.
New generation analytics and cognitive applications simply can’t be handled by traditional shared nothing Hadoop clusters, HDFS, and commodity storage. Those systems are designed for scale and low-cost, but don’t have the bandwidth or performance to support real-time analytics and cognitive systems that businesses collect and need to analyze data instantaneously. The new DS8880F data systems are designed specifically to handle these types of workloads that require lower latency and sub-millisecond response time.
The new DS8880F models also enable the consolidation of all mission-critical workloads, both new and traditional, for IBM z Systems and IBM Power Systems under a single all-flash storage family.
- Cognitive – Including Watson Explorer, Watson Content analytics and Watson APIs that allow customers and ISV’s to create their own Watson- based applications.
- Analytics- Including IBM Cognos Analytics, IBM SPSS, IBM Infosphere Big Insights, SAS business intelligence, Elasticsearch, Apache Solr and others.
- Traditional/Database – Including IBM DB2, Oracle, SAP, PostgreSQL, MongoDB, Cassandra and others.
DS8880F all-flash data systems –A closer look
IBM’s new line-up of DS8880F data systems includes three all-flash arrays designed to support applications from entry level business class to enterprise class to the “analytics” class required for real-time analytics and cognitive workloads. IBM uses a different approach in the design of its all-flash arrays. Rather than just replacing HDD’s with Flash cards IBM has completely rearchitected the system, optimizing it for Flash for better performance. The systems use the Flash Enclosure Gen2 with 2.5” drives in a 4U enclosure and are built on the Power CEC’s (Central Electronic Complex) used in Power servers. The models vary based on number of cores, capacity, cache and number of FibreChannel /FICON ports, but the functionality is the same across the product line.
Here are the details:
- DS8884F Business Class
- Built with IBM Power Systems S822
- 6-core POWER8 processor per S822
- 256 GB Cache (DRAM)
- 32 Fibre channel/FICON ports
- 6.4TB to 154 TB of flash capacity
- DS8886F Enterprise Class
- Built with IBM Power Systems S824
- 24-core POWER8 processor per S824
- 2 TB Cache (DRAM)
- 128 Fibre channel/FICON ports
- 6.4TB to 614.4 TB of flash capacity
- DS8888F Analytics Class
- Built with IBM Power Systems E850
- 48-core POWER8 processor per E850
- 2 TB Cache (DRAM)
- 128 Fibre channel/FICON ports
- 6.4TB to 1.22 PB of flash capacity
The Flash Enclosure Gen 2 provides significant improvements over the previous generation for both IOPs (read: 500,000+47%, write: 300,000+50%) and throughput (read: 14 GB/s + 268%, write: 10.5GB/s +288%). Aside from the performance gains seen from using all-flash, the unique architectural design attaches the enclosure to the PCI 3 bus rather than through the device adapter, eliminating a portion of the data path which lowers latency, and improves response time. The drawer itself has also been redesigned to provide higher bandwidth into the I/O drawer and to use ASIC rather than FPGA in the drawer itself (also improving bandwidth).
Other features of the DS8880F series include greater than “six-nines” availability; point-in-time copy functions with IBM FlashCopy; and Remote Mirror and Copy functions with Metro Mirror, Global Copy, Global Mirror, Metro/Global Mirror, IBM z/OS/Global Mirror, and z/OS Metro/Global Mirror providing disaster recovery/replication for up to 4 sites and 1000 miles with 3-5 second RPO (recovery point objective) for businesses where continuous operation is critical.
With this announcement, IBM expands its offerings of high-end enterprise-class storage by adding a new lower cost entry point that makes the systems affordable for a broader range of customers, while also enabling new uses cases that require the performance of all-flash. The optimized architecture further reduces latency and response time which will differentiate IBM in markets that require real-time analysis and results. Not only will these systems improve performance in their traditional database and OLTP realm, but the DS8880F family will become a key building block in infrastructure built to support cognitive and analytics applications.
In order to attract new customers, IBM also plans to expand marketing outreach by promoting the DS8880 systems as part of its broader cognitive computing strategy. The recent analyst briefing was a good start. Like IBM, I have been guilty of shunting aside the DS8000 family in favor of covering IBM Storwize, Cleversafe object storage and the IBM Spectrum software-defined family. In the future, I will pay closer attention. And IBM customers and prospects – you should, too.