Get to know the Zune HD>>
March 12, 2000 11:09 AM PST
Disclosure
Peter Glaskowsky has no business relations, investments or affiliations with subjects he covers except where noted in each post.
Peter N. Glaskowsky is a computer architect in Silicon Valley and a technology analyst for the Envisioneering Group. He has designed chip- and board-level products in the defense and computer industries, managed design teams, and served as editor in chief of the industry newsletter "Microprocessor Report." He is a member of the CNET Blog Network and is not an employee of CNET. Disclosure.
I worked for 6 years at Amdahl on their Unix product. As such I was exposed to the mainframe MP concepts and operation. My work was technical support. I had a few cases that dealt with MP issues in the kernel. I also had a boss that had been a field engineer and loved to give us glimpses into the hardware and machine OS concepts and operation. (The glass house/PC equivalent layer is the BIOS.)
I've been skeptical of the glass house efforts at MP. Mainly because they seem to be saying, "Hey, look what we created!" and seem to have no knowledge of the 3 decades of mainframe work at MP. I sometimes wonder if they avoid mentioning mainframe concepts or ideas for fear of being seen as a kin to mainframes.
Is my bias blinding me to a deeper aspect of the glass/house/PC market that actually does understand what mainframes achieved and is using that experience to advance the PC market?
Or am I correct in that the PC market is re-inventing the wheel out of ignorance or some belief that they are smarter or that the concepts and technology don't apply to PCs?
Or is it a more likely a complex combination of the two extremes and intermediate points?
I'd appreciate either a personal response or an analysis on this view.
Thanks!
JGT
John G. Thompson
The software always lacks behind, but so many people are trying to create a parallel processing software platform, that it will be done. It won't be created at one of the big corporations, it will be created by a little outfit, who will sell it to a larger corp, maybe a Cisco.
In fact, if Cisco owned a parallel processing platform, they would own the Internet backbone, the new software stack, the Cloud and everything in between. That would include all clients, and SaS service fees. They were the first that was suppose to take down Microsoft, and they are now in the best position to do so.
Virtualization is a Mainframe technology, cute but costly in computer overhead resources.
I read a 'white paper' this weekend published by Cisco, that discussed the cloud. It was a joint VMWARE, CISCO & EMC cloud.
It was crazy how much overhead it took to run a mere 60 VMWARE Virtual Machine environment.
Basically, since Virtualization can not provide any improvement in speed or performance, Cisco & EMC have rigged a ridiculous amount of hardware to create basically a virtual computer. And this virtual computer runs 60 VMWARE virtual machines. Performance, each machine can produce a mere 5 IOPS. That is 300 IOPS for 60 VM's.
The EMC storage server has, if I remember correctly, 280 15,000 rpm, 300 GB' hard drives and 16 Intel quad cores.
There are so many Cisco products involved it would be silly to list them here. But an example of one is that this Virtual Computer uses static load balancing and Cisco fibre to execute I/O executions to the massive memory machine provided by EMC. And the result once again, 5 IOPS per VM WARE VM's.
This my friends will not be our cloud computing model of the future. But in the interim it will make a whole lot of money for Cisco, EMC and VMWARE.
Tilera, Intel and multiple core offerings are binding their time until a Parallel Processing Platform, complete system shows up. Not Windows 7, Apple Snow Leopard, or any Linux or Unix offering. An entirely new platform will take us there.
And Amdahl's law is what they will need to overcome, capable multi tasks scheduling, that will get us beyond that third core, 12th thread. The threshold that stalls all.
Until then, the most cores anyone should ever purchase is 3, or a quad, (there is no three core offering). Because there is no gain in performance, and idle cores for low energy consumption is an embarrassing proposition to any respectable hardware or software engineer.
This is a historical time, we are stuck when it comes to performance, and as this and many articles have suggested, only software can lead us out of this technical quicksand.