"Fundamentally, computers have not changed since the 1960s, but the processes which bring technology together have changed greatly," said Jean-Pierre Garbani, principal analyst at Forrester Research of Forrester Research of Cambridge, Mass. "While applications cost millions before, you can now buy office tools for around $60."
Another big change, he said, is how IT is now an integral part of the business as a whole. No longer is it relegated to the back office. Technology and related processes are so woven into the fabric of the company that there is no hiding place for IT in some ivory silo that management rarely visits. Thus IT is very much on the radar screen of top management, and that's why it appears to be getting in more trouble than before.
This trend, he said, has been even further accelerated by such developments such as HP BladeSystem Matrix and Cisco Unified Computing System (UCS).
"HP BladeSystem Matrix and Cisco UCS will bring about a deep change in IT," said Garbani, "as they are accelerating the trend of IT services becoming a key aspect of business productivity. This dependency means that IT is part of the cost of doing business and part of the bottom line."
He described how the volume and complexity of technology is advancing in the enterprise, driven by Moore's Law. As the capabilities of hardware increase not just the chips, but also memory, I/O, blades and more software is filling the vacuum. Thus, better management tools are needed.
A few years ago, companies had to invest in expensive software implementations like UniCenter and Tivoli. These days, everything is being rolled into the platforms themselves. UCS, for example, already has BMC management software built right into it. This provides plenty of management functionality inside.
"UCS has BMC change management, provisioning and configuration management database (CMDB) features inside," said Garbani. "In many ways, what is happening here is the re-creation of the mainframe in terms of efficiency."
What is missing in these new platforms, he said, is capacity management capabilities.
"These new compute platforms, characterized as a cloud in a box, will change the world," said Garbani. "But they require sophisticated capacity planning tools, such as TeamQuest, to do the whole job."
Without that capability, virtualization and hardware consolidation projects are largely going blind. Garbani gives an example of a common mistake when trying to translate between the physical and the virtual world an administrator takes a 2 GHz processor and divides it up among five virtual machines, with each one allocated 400 MHz. That doesn't work, as it doesn't take into account the fact that instructions have to go via the Guest OS, hypervisor, OS and the hardware, thus consuming overhead.
"The execution time is not the same with virtual CPUs and there is no direct linear relationship between the virtual and the physical in terms of performance," said Garbani. "The only way to be sure is to model well using capacity planning tools."
Bottom line: it might be easy to define everything in an ideal virtual world, but the reality is that it is a lot harder to execute.
"We used to talk a lot about resource contention back in the sixties and this is making a comeback," said Garbani. "If you have one Network Interface Card [NIC] in a physical box and set up 4 VMs on that box with 4 virtual NICs, you are going to have contention problems."
He offered an automotive analogy as an explanation. Hardware people give you a very fancy car with the most powerful engine, cornering and gearing capabilities. But that vehicle isn't any better than a cheap model if you don't know how to drive it and how to exploit its capabilities.
"Hardware gives you the car but doesn't teach you how to drive or where to go," said Garbani. "The best driving school is capacity management."
Article courtesy of Server Watch.