Hewlett-Packard's new x86 server line, announced Monday, aims to wring out as much management overhead as possible through automation and improved energy management.
HP today provided an overview of the ProLiant Gen8 server plan, including a look at features and capabilities.
Details on specific ProLiant Gen8 servers and their price tags will come later.
For now, HP wants its customers to know that it spent some $300 million to develop the ProLiant Gen8 servers under a program it called "Project Voyager" -- and filed 900 patents related to the effort.
The patents and investment are big numbers, though to put the announcement in context, HP is routinely one of the top patent recipients in the U.S. The company was granted 1,480 patents in 2010alone, ranking it 10th overall that year.
And spending $300 million over two years on Project Voyager is probably affordable for a company that reported revenue of $127.2 billion last year alone.
Nonetheless, HP said the new server line has has built-in functionality, such as separate memory on the motherboard to house deployment and provisioning information to reduce system administration overhead.
The servers can monitor some 1,600 parameters, information that's automatically linked with HP services to speed response to problems and even automatically ship replacement parts without human intervention.
The monitoring is controlled by a secondary processor in what HP calls its Integrated Lights-Out (iLO) management engine. The servers and racks also have location sensing.
HP said it did a lot with storage performance, using three-drive mirroring to improve read and write performance, as well as optimizing its storage subsystem and application performance by using solid state disk drives.
"Clearly, we're not just launching a commodity server," said Jeff Carlat, director of product marketing for HP's Industry Standard Servers and Software.
ProLiant Gen8 beta tester Purdue University built a supercomputer cluster with the servers using Intel's latest Xeon chip, code-named Sandy Bridge. The so-called Carter Cluster is named for Dennis Carter, a Purdue alumnus, electrical engineer, and retired Intel vice president credited with developing the "Intel Inside" marketing campaign.
The Purdue system is ranked 54th on the Top 500 list of supercomputers, with 10,368 compute cores running at 187 teraflops.
Purdue builds a new cluster machine each year. Gerry McCartney, CIO at Purdue, said the latest machine draws less power than any of the previous four machines, "which is quite remarkable."
McCartney credits the power saving features in the HP servers, particularly the ability to dynamically slide power consumption as demand is placed on the machine for lower energy consumption.
"Power is the constraining factor now in high performance computing," said McCartney.
HP also said it increased the number of sensors in the new servers. Previous servers had sensors across the motherboard, but it not also up and down along the risers like the new ones.
Carlat says the system will deliver 1.7 times more compute power per watt against Gen6, a three-year-old server and the one he believes customers are most likely to upgrade from.
The cloud management capabilities in ProLiant Gen8 provide a degree of automation, but aren't yet offering infrastructure management capabilities called for in an HP server roadmap, said Carlat.
McCartney welcomes cloud management tools. "The days of when you have to run your own hardware -- if they haven't already gone, they're about to go," he said.
While HP has put pieces of its sensor and management technology in its servers and racks in earlier generations, it says this server release ties it all together it what it calls the "ProActive Insight architecture."
Jonathan Eunice, an analyst at Illuminata, said HP's previous approach was to make dozens of incremental changes to its servers. But with the new server generation, "the ambition is very grand compared to previous releases," particularly the locational capabilities and energy management, he said.
Eunice said that HP is responding to competition coming from new server entrants, such as Cisco, as well as from cloud providers. HP is "doing things in a manageability space that you can't just do on the virtual level," he said.
"They are really kind of building the infrastructure for a smart data center," said Eunice, who likened the effort to a longtime industry goal of autonomic computing or self-managing functionality in a distributed environment.
Patrick Thibodeau covers SaaS and enterprise applications, outsourcing, government IT policies, data centers and IT workforce issues for Computerworld. Follow Patrick on Twitter at @DCgov, or subscribe to Patrick's RSS feed . His e-mail address is email@example.com.