September 17, 2008
Like death and taxes, the perplexities of data integration are unavoidable. The never-ending data explosion is just one factor. Mergers and acquisitions, globalization, outsourcing, partnerships, and regulations all contribute to the massive pile-up of data from different repositories, business systems and applications, in different formats and languages, structured and unstructured. Giving people a single, cohesive view into what would otherwise be a massive quagmire is not easy.
“Data integration is always a challenge and will remain that way because data grows exponentially and new types always have to be added to the mix,” says Krishna Roy, enterprise software analyst for The 451 Group. “Keeping up with the growth of data and enabling it to be not only integrated, but cleansed, as well, is the big challenge.”
According to Roy and other enterprise IT analysts, one of the leading companies meeting that challenge is Informatica. The company's flagship, PowerCenter, is a platform for accessing and integrating data from different business systems and repositories, then sharing that data throughout the enterprise. A grid version lets organizations distribute data integration tasks in a scalable, resilient, high-performance environment.
PowerCenter, like Informatica, has evolved over the years to handle all the different chores involved in an integration undertaking, including data migration and replication, synchronization, master data management, governance and standardization. "We started out to help with the automation of data warehouses, taking data from lots of different sources and providing a holistic view of it, whether it came from mainframes, packaged applications, databases, message queues, all the different feeds, and even data outside the enterprise," says Adam Wilson, senior vice president of product management and marketing. During the past eight or so years, the company has expanded beyond data warehousing into “broader data integration, including data from outside the enterprise, data that’s structured or unstructured,” Wilson says. “It’s really broader business intelligence we’re focusing on.”
The increasing need for better data integration is being driven, Wilson says, by companies trying to "get proactive about governing their data, and providing access to that data,” plus globalization, which brings not only new data systems to contend with, but also new sources, such as partners and providers, not to mention new formats, character sets and regulations.
After Virgin Media in the United Kingdom acquired ntl and Telewest as part of its expansion into online and telephony services, the firm had 20 new data sources and 5.6 million customers to consolidate into their operational systems. "They said they really needed a way to pull together all this data that's smeared across all these different systems,” Wilson says. “They wanted a single view of all their 10 million customers, and that information was kept in an Oracle data warehouse, various customer management systems, and running on a mix of hardware and operating systems.” Taking advantage of PowerCenter’s real-time capabilities, Virgin was able to build a consolidated data integration hub that delivers updated information, resulting in better customer service and more accurate market analysis.
Getting on the Grid
With PowerCenter 8 in 2006, Informatica extended its integration capabilities to the grid, enabling customers to distribute tasks across multiple processor nodes while taking advantage of commodity hardware, scalability and high availability. With the Enterprise Grid Option, Informatica says, it has developed a grid system that understands data integration tasks and the resources they need, and is able to adapt accordingly. PowerCenter 7 -- way back in 2003 -- included some basic grid support, but the version 8 and the Enterprise Grid Option brings the platform's whole shebang to a grid environment: universal access to a wide range of applications and legacy systems; data delivery functions; and tools for capturing, cleansing, managing and migrating data.
In the PowerCenter grid, data integration services, repository services, and logging services run on one more nodes (logical representations of a physical machine). On each node, a service manager handles the services assigned to that node. Each manager keeps statistics on available CPU usage, available memory and number of running threads. What Informatica calls “gateway nodes” control the routing of service manager requests. The gateway keeps an eye on the availability of other nodes, manages application services, and makes sure services execute by dynamically redirecting to a secondary node if the primary node is down. Administrators can specify the resources available to run a task on each node, and also assign higher priority to the most important integration operations. A GUI Web console provides central control for adding nodes or services and managing resources across the grid.
Adaptive load balancing dynamically assigns and executes tasks on the basis of resource availability or according to the resource requirements of a particular data integration task. Dynamic partitioning adjusts the parallel execution plan when nodes are added or dropped. PowerCenter's load balancing technology is platform-agnostic and can interoperate across a heterogeneous grid environment (Linux, Windows, Unix; different CPU speeds; 32- and 64-bit software; varying memory capacities among nodes; etc.).
To guarantee integration tasks are processed according to business priorities, PowerCenter uses resource reservation to set aside certain nodes to perform certain tasks, so that a CPU-intensive job, for example, would go to a node with the most CPU power and memory; the administrator can then make sure that node gets the job. Admins can assign different service level values to tasks, based on priority. The "gang dispatcher" makes sure that when tasks are broken up into subtasks to process on different nodes, all subtasks are executed at the same time. Resource threshold provisioning is meant to help avoid overloading of nodes by controlling the number of threads, amount of memory that can be consumed, and the number of concurrent data integration tasks; the company says this allows load spikes to be processed without degrading overall performance.
“We have created a data services platform that can deploy data integration, metadata, and profiling and connectivity services out in a grid,” Wilson says. “Customers typically use us in one of two scenarios. We’re moving terabytes and terabytes in incredibly short load windows, so we break the work down, parallelize it, and spread it across the grid, using rules-based or cost-based optimization. In the other scenario, customers have built data services within our Informatica software and they are calling us in a request-and-response manner, where we have to handle thousands of requests for those data services, so we scale up to distribute those requests to nodes that have been identified in our domain.”
“In both throughput and concurrency situations, our ability to distribute workloads gracefully across a grid is our way to ensure that customers’ service level agreements are honored,” Wilson says. “That’s why we’ve incorporated things like automatic load balancing and resiliency, and things like letting every task have its own service level based on its business priority.”
Millions of Transactions a Day
LinkShare, a Web marketing company with thousands of clients, needs to track performance, order transactions, lead generation, clicks on ads, content inventory, and other items. As the company has acquired customers and partners, data volumes have grown to the point that it is capturing information from more than 400 sources and processing more than 300 million aggregate transactions a day. PowerCenter, which LinkShare has been using since 2001, extracts and transforms all this transaction data, puts it into a near real-time data store, then extracts it again into an enterprise data warehouse. PowerCenter provides the processes that enable the different LinkShare systems (for invoice processing, order processing, payments, reporting, analytics, etc.) and applications (MySQL, Oracle 10g, IBM DB2) to talk to each other.
Customers are feeding data into the system around the clock, so LinkShare needs to ensure high-performance processing of large data volumes in order to meet service level agreements. The company taps PowerCenter to load-balance more than 700 sessions that run in short batch-processing windows throughout the day.
LinkShare had been a PowerCenter user since 2001, but in 2007 moved its operation to the Enterprise Grid Option while also moving to 64-bit hardware. With their previous cluster, the IT team had to manually configure load balancing and resource utilization. The Enterprise Grid Option takes care of those functions dynamically, without any re-coding of applications, Wilson says.
'A Good Place to Be'
By exploiting grid architecture and its own platform and technologies, Informatica is able to deliver the capabilities required for real-time processing of data integration tasks. “The leading challenge to data integration practitioners today is the need to move data in real time or on demand,” say Philip Russom, senior manager of The Data Warehouse Institute’s TDWI Research wing. “Most data integration tools and user practices were originally designed for latent batch processing in nightly windows. Vendor tools have come a long way in this regard, and users have educated themselves in more dynamic applications of data integration.”
“Informatica is one of the leading vendors in the data integration space, along with Business Objects, IBM, Microsoft, Oracle and SAS. Informatica has shown leadership in recent years by championing new practices and techniques like data integration competency centers, service-oriented architecture, embedded data quality functions, and business-to-business data integration,” Russom says.
The latest PowerCenter 8.6 "sports a number of enhancements that form part of an ongoing mission to extend its usage beyond extract/transform/load (ETL) projects for data warehousing," says The 451 Group’s Roy. "The advent of the Real-Time Edition, the first tools for analysts and data stewards, a new B2B Data Transformation engine and a data-loader service for integrating off-premise data managed by Salesforce.com with on-premise data, are some of the improvements in 8.6 crafted with this mission in mind."
The company continues to "retain the mantle of largest independent data integration vendor," according to a report issued in the August by The 451 Group (while noting its competitors include IBM and Oracle, rivals with whom it also shares technology). "Informatica has been increasing revenue by 20 percent year-on-year for several years now -- a growth rate it maintained for the first six months of 2008 when sales rose to $217.5 million from $181.4 million in the same six months of 2007."
According to Informatica's Wilson, the company's installation base "is growing 20 percent year over year while enterprise software is growing 8 to 9 percent. Ninety of the Fortune 100 companies are our clients." The biggest problem Informatica solves today is "ETL into the warehouse," Roy says. "However, data migration, compliance/data governance, and master data management are all areas in which Informatica increasingly plays. I think being the Switzerland of the data integration world is a good place to be given that integration software needs to be independent of databases, applications, and so on."
10/30/2013 | Cray, DDN, Mellanox, NetApp, ScaleMP, Supermicro, Xyratex | Creating data is easy… the challenge is getting it to the right place to make use of it. This paper discusses fresh solutions that can directly increase I/O efficiency, and the applications of these solutions to current, and new technology infrastructures.
10/01/2013 | IBM | A new trend is developing in the HPC space that is also affecting enterprise computing productivity with the arrival of “ultra-dense” hyper-scale servers.
Ken Claffey, SVP and General Manager at Xyratex, presents ClusterStor at the Vendor Showdown at ISC13 in Leipzig, Germany.
Join HPCwire Editor Nicole Hemsoth and Dr. David Bader from Georgia Tech as they take center stage on opening night at Atlanta's first Big Data Kick Off Week, filmed in front of a live audience. Nicole and David look at the evolution of HPC, today's big data challenges, discuss real world solutions, and reveal their predictions. Exactly what does the future holds for HPC?