Hewlett-Packard's emergence as a player in data warehousing (DW) is one of the year's most important developments in information management. Until current CEO Mark Hurd assumed leadership, HP's presence had been limited, despite the fact that the company regularly is at or near the top of the market in servers and other infrastructure technologies that support BI and DW implementations.

Aiming for what it describes as the gap between departmental data marts using DW appliances at the low end and very large enterprise data warehouses at the high end, HP has launched version 2.0 of NeoView, which offers " black box" configurations of database software, hardware and networking. Ventana Research advises companies that are interested in the DW appliance approach but that are looking to support mixed query workloads to examine NeoView. Looking forward to 2008, HP intends to employ parallel blade servers and 10-gigabit Ethernet over InfiniBand as part of the NeoView configurations. These two advances should lower the cost of scaling up to handle large, mixed BI and analytical query workloads, which today require mainframes or other high-end, centralized servers.

The Case for Consolidation

Under Hurd's leadership, HP has aggressively applied BI and DW to its own internal management practices and has brought new DW products and services to market. HP is a business behemoth; its US$57 billion supply chain ships three printers and two PCs, among other products, every second. HP has more than 1 billion customers and records 50 million line-item orders per year. To overcome its expensive and frustrating complexity, Hurd is making HP the leading case study example of how to apply its advanced BI and DW technology.

Like many large organizations, HP is trying to consolidate data centers and applications to cut costs and to pave the way for operational BI, which Ventana Research sees as critical to improving business performance by enabling line employees to execute on strategy. While creating a multi-terabyte enterprise data warehouse, as HP is trying to do, is not at all simple, today's typical environment of disparate, heterogeneous data sources carries its own significant cost.

" If you stick 700 data marts and 5,000 applications in 87 data centers and run queries against all this siloed data, you quickly develop the world's largest telecom bill," noted Hurd in a recent speech. " When I became CEO, I knew I was in trouble when the heads of all the telecom companies called to congratulate me and say I was their best customer."

A Closer Look at NeoView

NeoView is the foundation of HP's enterprise DW offering. A preconfigured system that combines hardware, software and an interconnecting " fabric," NeoView offers a " shared nothing" architecture, which means that nodes in its massively parallel processing (MPP) system are independent and do not share memory or storage. In NeoView, the nodes are tied together by ServerNet, a proprietary interconnect technology developed for Tandem Computers, which HP acquired as part of Compaq in 2002. HP plans to replace ServerNet eventually with 10G Ethernet when that technology's performance is sufficient.

NeoView derives much of its current technology from HP's ongoing redevelopment of NonStop SQL, a largely dormant product that Tandem originally designed as an analytical database to run on Microsoft Windows NT clusters. With the NeoView 2.0 release, HP is taking its significant re-engineering investment to market in the hope of catching the next generation of BI, DW and analytics implementations, which will focus on operational users and tasks. These implementations will demand far greater scalability, availability and flexibility than was required of past implementations. In NeoView, HP also brings to market a new query optimizer and a " performance repository" feature to tune frequently used data for production reporting and other BI uses.

DW Appliances in Perspective

Data warehousing appliances are becoming a critical component of many organizations' information management strategies to extend BI and analytics to operational users and varied business processes. Many companies are keenly interested in alternatives to the traditionally costly and time-consuming task of configuring and tuning database, server and storage systems for data mart and data warehouse implementations. HP's NeoView, like most DW appliances, employs massively parallel processing (MPP) technology, which scales easily and relatively inexpensively to handle CPU-intensive queries and table scans. Configuration pricing is straightforward, based on the number of processors and terabytes of storage needed.

NeoView differs from appliance competitors like DATAllegro and Netezza in that it serves the needs of organizations – like HP itself – that are intent on consolidating data marts into an enterprise data warehouse. These users need a system that supports mixed workloads of different types of BI and analytical queries. Ultimately, this positioning will put HP in direct competition with Teradata, which specializes in DW, has a parallel system, and can handle mixed BI and analytical workloads running at a very high level. In the short term, NeoView probably will be of most interest to companies that are struggling to scale up traditional DWs running on non-MPP databases but are hesitant to incur the greater expense and configuration complexity these implementations require.

Recommendation

Ventana Research recommends that companies that plan to consolidate data marts and establish an enterprise DW take a serious look at HP's NeoView. While the database and DW industry is well-established, changes in demand – especially for operational BI – will shake things up and should give organizations cause to re-examine their data warehousing and information management investments. Along with their benefits, DW appliances have a potential downside in that they could exacerbate the information management problem of out-of-control data marts. Once it has matured, however, HP's approach could offer customers the best of both worlds.

David Stodder is vice president and research director - information management and IT performance management at Ventana Research. He was previously editor-in-chief and editorial director of Intelligent Enterprise.