IBM has expanded its family of PureSystems by adding a whole new box: IBM's PureData System. It's designed, as you might guess, to help enterprises analyze petabytes of data for marketing, sales, and business operations. According to IBM, 2.5 exabytes of data are created every day, which is so much data that 90 percent of the world's data has been created in the previous two years.
IBM's new PureData System is available in three workload-specific models optimized for transactional, operational, and big data analytics.
But IBM's PureData Systems is a marker of something more important—IBM's overall strategy to deliver a family of systems designed to reduce the complexity of implementing technology. PureData System joins IBM's PureFlex System and PureApplication System, both of which were introduced early this year. The basic idea is greater flexibility, more computing, and much faster implementation and setup times.
Will PureSystems replace single servers like the traditional IBM Power Systems anytime soon? Not likely, because the cost-to-value of more traditional IBM Power Systems are already great. Besides, for many of IBM's existing customers, their existing infrastructures that utilize IBM Power Systems aren't broken, so they're not going to get "fixed."
Growing the Pure Lineup
PureSystems are a new class of density-optimized servers that are essentially a modern reboot of the blade-based system. Density-optimized systems from IBM (and its competitors) have more flexible chassis designs that let customers drop in very powerful nodes (instead of blades) and still end up with a discrete footprint in the datacenter. But IBM has a far bigger vision for PureSystems, because PureSystems are all about entire classes of quickly installable solutions—see IBM's PureSystems Centre
—as well as pre-configured hardware setups to handle specialized tasks. And the latest case in point is PureData System.
As it turns out, PureData System is actually available in three workload-specific models:
- PureData System for Transactions. Think retail and credit card processing environments with high volumes of transactions that might also need to scale rapidly during certain seasons.
- PureData System for Analytics. Think big data in the land of petabytes. Powered by Netezza technology, this system simplifies and optimizes performance of data warehouse services and analytics applications. Clients can use it to predict and help avoid customer churn, create targeted advertising and promotions using predictive and spatial analysis, and prevent fraud.
- PureData System for Operational Analytics. Think about real-time action for fraud detection during credit card processing, to deliver customer insights to call center operations, and the ability to track and predict real-time changes in supply and demand for energy and utilities.
Although these systems might seem too big and complicated for the low end of IBM's Power Systems customer set, IBM is clearly resetting expectations for server systems for medium-to-large enterprises. For example, the Premier healthcare alliance has selected a PureData System to manage and analyze a large clinical, financial, and outcomes database that has information on one in every four patient discharges in the U.S., 2.5 million real-time clinical transactions a day, and $43 billion in annual purchasing data, IBM reports.
“One of our biggest challenges has been the ability to quickly acquire and analyze big data, then place the insights from that data directly into the hands of caregivers,” explained Todd Wilkes, vice president of Enterprise Solution Development at Premier, a provider-owned alliance helping 2,700 hospitals and 90,000 other health providers improve their patient care and finances.
“With IBM big data analytics, we can now analyze any data with sub-second response times and share insights from it to help our alliance members improve patient care, reduce costs and better understand what’s going on within their hospitals and health systems.”
IBM's PureFlex System is aimed at creating and managing an IT infrastructure, whereas the PureApplication System is for reducing the cost and complexity of rapidly deploying and managing applications.
With a PureFlex System, IBM lets customer customize it with what IBM calls Open Choice. Customers can configure the system with POWER-based nodes running AIX, IBM i, or Linux, for example, in addition to Intel-based nodes, and even customize the hypervisor and networking.
PureApplication System is similarly transformable based on what IBM calls Patterns. There are three basic kinds of patterns:
- Infrastructure Patterns. Think automated, policy driven infrastructure management across each node, including storage and networking.
- Platform Patterns. Think pre-configured policy managed services for caching, elasticity, failover, load balancing, security, database, and middleware.
- Application Patterns. Think pre-defined application architectures and required platform services, deployed and managed through policies.
These patterns are designed to make it much easier to implement systems that have application and infrastructure dependencies that might not fit in a single traditional server, yet with a PureSystems, you can setup and manage all of these sorts of resources in the same physical location, as well as manage the whole package with a single management and monitoring console.
In late October at the IBM Power Systems Technical University event in Las Vegas, Pat O'Rourke, a senior consultant for client advocacy and system assurance for the IBM Systems and Technology Group, pointed out the sort of flexible scalability found in a PureSystems solution. To run POWER-based workloads, customers can choose between the IBM Flex System p260 or p460 Compute Nodes. In terms of performance, the p260 compares to IBM's Power 710-to-740 line of Power Systems, O'Rourke explained, noting that the p460 relates well to the Power 750.
"Imagine having the ability to put 14 740s in a chassis that is 10U in height—terrific consolidation," he said, adding that essentially the computing power of seven Power 750s (using p460s) could fit into a PureSystem. Plus, if a customer needed to, the available nodes could be co-populated with x86-based nodes.
It’s this kind of scalability and flexibility that has IBM so excited about PureSystems.