But Wait, There's More
Linux Gets Native InfiniBand I/O Support
OpenIB Alliance, the industry association formed last summer to promote the InfiniBand I/O architecture and an open software stack that runs on top of InfiniBand devices, announced last week that the OpenIB software stack and core drivers have been incorporated into the last set of patches to the Linux kernel. While InfiniBand was heralded a number of years ago as a means of creating seamless server and storage switched fabrics, the InfiniBand protocol has lost its initial traction--big surprise--in the high-performance-computing space as a means of lashing together nodes in massively parallel Linux clusters.
The OpenIB Alliance members say they will now focus on developing upper-level InfiniBand protocols and APIs. The organization will be hosting a developers' workshop from February 6 through 9 in Sonoma, California, to hammer out the proposed OpenIB software architecture, most notably including Remote Direct Memory Access (RDMA) over InfiniBand, which would link the main memories of nodes in a cluster directly to each other--rather than passing through cache and I/O buses--and thereby more tightly coupling nodes together while reducing latencies in data transfers between nodes.
IBM Contributes to Xen Open Source Virtualization Project
It's only been a few weeks since the Xen open source virtual machine project burst onto the commercial scene, and already IT industry stalwarts are lining up to help out on the project. Reiner Sailer, who works at IBM's T.J. Watson Research Center, in Yorktown Heights, New York, put up a posting on SourceForge explaining that he worked for IBM's secure systems department and had developed a security architecture for hypervisors, called sHype, which IBM has created for its own X86-based hypervisor.
First, IBM was working on its own hypervisor? Hmm . . . very interesting.
Second, having created sHype for this research hypervisor for X86 machines, IBM has decided to contribute the sHype code to the Xen community so it can be integrated into this up-and-coming virtual machine partitioning environment. Sailer says that sHype is itself based, in concept, on the Flask security architecture of the Security Enhanced Linux (SE-Linux) variant of the Linux kernel for ruggedized applications. IBM's goal is to harden Xen to the point that it can get Common Criteria certification, which makes it suitable for financial services and defense applications.
Judge Tells IBM to Stop Dragging Its Feet and Cough Up the AIX and Dynix Code for SCO Suit
The SCO Group has scored a minor victory in its $3 billion Linux-Unix lawsuit with IBM, with the U.S. Magistrate Judge on the case, Brooke Wells, issuing an order that compels IBM to respond to further discovery by SCO's lawyers and to make available more source code for the AIX and Dynix Unix variants that IBM controls. By opening the discovery process a little wider, SCO will presumably be able to find the smoking gun that proves IBM illegally dumped Unix code into Linux. This presumption may, of course, turn out to be just that: a presumption, not a fact. IBM has been given until March 18 to produce the additional code and related documents. Provided the summary motions to dismiss that IBM filed are not effective, the case will go to trial sometime in November. But you can bet IBM's lawyers will continue to push the legal limits of time-stretching, which is why SCO cleverly negotiated a cap on legal fees with its lawyers. By the way, if IBM doesn't cough up the code, Judge Wells has said she will then order SCO to have unfettered access to IBM's code repositories.
In a separate development, two weeks ago, SCO's lawyers filed a motion with Judge Kimball for IBM's chairman, Sam Palmisano, to be deposed as part of the lawsuit. Back at the end of 1999, when IBM first caught Linux fever, it was Palmisano, as head of IBM's Server Group, who lead the charge. Shortly thereafter, he was named president and chief operating officer, thereby being anointed as the company's future chairman and CEO once Louis Gerstner retired. IBM's lawyers, according to the SCO filing, have contended that, "Palmisano does not have 'any knowledge' of any facts relevant to this litigation." This seems unlikely, since he was the one who picked the Linux czar at IBM and realized that Linux was going to be a gravy train--at least if you believe the PR IBM was generating at the time.
Linux Racks Up More TPC Benchmark Numbers
Database maker and ERP juggernaut Oracle continues to spend big bucks on TPC performance benchmarks to prove the mettle of the Linux platform in server environments. Oracle is not the only one spending money on these tests, however: server maker Hewlett-Packard is helping out as well, to promote its Itanium and Opteron systems.
HP and Oracle have just announced a TPC-C benchmark test result on an Integrity rx4640 server, using four of the new 1.6 GHz, 9 MB cache "Madison" Itanium 2 chips. In that online transaction processing test, the server was equipped with 128 GB of main memory and 19.2 TB of disk capacity, and it could crank through 161,217 transactions per minute (TPM). That server was equipped with Red Hat Enterprise Linux AS 3 and Oracle's 10g Standard Edition. This machine delivered a price/performance of $3.94 per TPM. That performance is slightly better than for eight-way Xeon MP servers running Windows Server 2003 and delivering about the same bang for the buck at the network-level pricing that the TPC-C test uses. That Linux box did a little less work than a four-way Power5-based p5 570 server using 1.65 GHz Power5 cores. That AIX-based machine delivered 194,391 TPM, but did so at a cost of $5.62 per TPM--even after hefty discounts.
On the TPC-H data warehousing benchmark test, a 12-node cluster comprised of four-way ProLiant DL585 servers with 2.2 GHz Opteron 848 processors was able to handle 35,141 queries per hour (QPH) on the 1,000 GB version of the test. Those servers, which were also running Red Hat AS 3 and Oracle 10g Enterprise Edition (which includes the clustered database software), were clustered together with InfiniBand switches and host software from InfiniCon Systems. The cluster costs $60 per QPH, which was about the same bang for the buck as UltraSparc and Itanium servers that delivered a lot less performance (but could have scaled, too, if they had been extended with clusters). The performance of this Linux cluster was a little bit better than a monolithic Fujitsu-Siemens PrimePower 2500 Unix server, which cost 2.5 times as much. An HP Superdome using 64 of Intel's 1.5 GHz Itaniums could deliver 45,248 QPH on the TPC-H test, but it cost $109 per QPH, too.
Wary CIOs to Spend 2.5 % More on IT in 2005, Says Gartner
The analysts at Gartner have dusted off their crystal balls and finally joined the shrinking crowd of analysts (last year Forrester bought GIGA and Gartner bought META Group) in prognosticating about IT spending growth in 2005. According to a survey of 1,300 chief information officers, who have a combined spending budget of $57 billion (or about $44 million each, on average), across a cross section of industries that is representative of the worldwide economy, IT budgets are expected to rise only 2.5 percent in 2005. This is considerably lower than all but the most pessimistic estimates for IT spending growth that came out as 2004 was coming to a close, which had growth pegged at somewhere between 5 and 7 percent.
The respondents of the Gartner survey showed why the budget numbers were lower than expected, and why that number is increasingly irrelevant. First and foremost, two-thirds of the CIOs polled say that there is a disconnect between what they believe and what their CEOs believe they can and need to do with IT systems, and these CIOs portrayed themselves as being "at risk" because of the way CEOs view the performance of their IT departments.
And while CIOs ranked business process engineering (BPO), beefing up security, cutting operating costs, supporting the competitive advantages of their business, and revenue growth as key business factors, the top IT priorities do not map one-to-one, with IT executives saying that they want to have better security tools, more business intelligence and analytics, wireless and mobile access to systems for the workforce, and ERP upgrades. There is a big disconnect here, and that does not bode well. The other disconnect, which companies like IBM and Accenture are hoping to exploit, is one between the skill sets for business process engineering (which involves a rethinking of the processes that get automated into systems to simplify business and streamline the applications that support those processes) that companies have in IT and other departments and the skills that IT directors believe they will need to do BPO. Some 61 percent of IT executives polled said they did not have the skills they needed in general, and of those who ranked BPO in the top five of priorities for this year, 80 percent said they didn't have the skills to actually do BPO. Yikes.
Survey Says IT Salaries Stagnated in 2004
IT Salary survey specialist Janco Associates, based in Park City, Utah, has just released the results of a salary survey for the second half of 2004, and the results seem to indicate that IT salaries in North America are still under pressure from the combined effects of weak economic conditions and offshoring.
We keep hearing how IT budgets grew in 2004 and are expected to grow again in 2005, but Janco's surveys suggest that average compensation across all IT jobs and business sizes that Janco surveys (midsized and large enterprises) actually fell by 0.75 percent from June 2004 to January 2005, to $77,841. For many job titles, including operators, analysts, customer service managers, Webmasters, office automation managers, graphics and forms designers, and project managers, average salaries fell by 5.1 to 13.2 percent over that time. Salaries for voice and wireless telecom managers fell by 15.2 percent, which is stunning. At large enterprises, salaries for IT administrators, Web application developers, data warehouse managers, LAN application managers, and other hot areas--where the average salaries tend to be larger--grew from 5.1 to 8.5 percent. Only nine of 24 non-executive positions tracked by Janco saw average salary growth, but the salaries that these top people command are large enough to raise the class average and to mute the declines across all titles. To put it bluntly, the IT industry average salary masks huge declines in salaries for key IT positions.
In the executive ranks, Janco says, total average compensation for CIOs at large enterprises was up by 4.2 percent in the six months, to $169,601. At midsized companies, the average salary was up only 1.4 percent, to $171,791. Those numbers are not typos. You see it right; on average, CIOs at midrange companies are getting paid a little more than those at large enterprises. That's because big companies have been cutting executive compensation (mostly by paying less for new people) since it peaked in 2000. Across all industries and job titles, Janco says, many people who had planned to retire in 2003 and 2004 have still not seen their retirement portfolios recover from the recession of 2000 to 2003, and they have opted to keep working and not push for big pay raises.
Gartner Says Utility Computing Will Cost IT Jobs
According to a new report from IT researcher Gartner, the advent of utility-style computing, with its automation of provisioning and systems management, will mean that heads will start to roll at external service providers as well as the internal IT organizations of corporations. This will not be welcome news to the system administrators and network operators who already work in stress-laden IT organizations that have been looking to reduce their IT administrative costs over the past four years.
The Gartner report, released this week in the United Kingdom, concludes that, over the next two to 10 years, data center employees will be increasingly replaced by the machines they love, bringing an ironic full circle to the automation that started with record-keeping and billing at the dawn of the Computer Age decades ago.
"The trend towards offshore services has monopolized attention in terms of job losses," said Gianluca Tramacere, an analyst in Gartner's IT services and sourcing group, in a statement accompanying the report. "There is less awareness that increasing reliance on highly automated infrastructures will significantly reduce the need for manual procedures and direct involvement of the workforce. IT automation can mean greater flexibility and cost efficiency for businesses. However, it makes it harder for IT personnel to defend their jobs as this evolution, accelerated by the global economy and the competitive marketplace, is seen as an inevitable consequence of IT progress."
The challenge for IT personnel, according to Gartner, will be for administrators and system managers to learn skills that are in demand, and that goes for those who work in the corporate data center as well as those in the giant services companies that run other companies' data centers. There is no escaping the automation of the computers. And Gartner's advice to IT personnel? Learn all you can about your company's business processes and become an expert in that. To a Unix or mainframe operator, this advice may seem somewhat bizarre, and the reality is that a company will not necessarily need as many business process experts as it needs system analysts and administrators. There will probably be a lot more people playing musical chairs in the data center than there are chairs to sit on.
The jobs issue is further complicated as companies start dreaming, thanks to the application service provider, utility, and subscription pricing models, which companies like IBM, Sun Microsystems, and Hewlett-Packard keep talking up. If companies rent rather than own their infrastructure, with the advanced networking and decent management tools available today it doesn't much matter where those machines are located or who babysits them. If you subscribe to your servers, which are a relatively small portion of your budget, why would you not subscribe to your system administrator, which is a big portion of your budget? Sharing people and spreading costs among many different companies is what we all do for installation and break-fix maintenance support today. This just takes it all one step further.
"ESPs, including IBM and HP, are investing heavily in new technologies that will allow the automated delivery of IT services," said Tramacere. "Examples of technologies that are emerging include 'self-healing' hardware, rapid development tools and software components, and tools that automatically manage systems and services. In the future, strengthened by the economies of scale, technical enhancements, and adaptable pricing, organizations will be able to access IT utility infrastructure services more competitively, with a reduced reliance on internal resources. Ultimately, organizations will be driven to access utility infrastructure services more frequently and to reduce the size of their internal IT operations."