RSSIBM 'Virtualization Engine' Maximizes Computing Power
By David Worthington | Published April 30, 2004, 6:05 AM
In an effort to simplify virtual computing, IBM has glued together tools previously scattered across its eServer lineup and developed a preview of its Virtualization Engine. The software reclaims and consolidates unused processing power and storage systems with the end result being purely economic - customers acquire more "bang for the buck."
Big Blue's Virtualization Engine could be considered a jack of all trades. The technology partitions a single CPU into up to ten independent servers, harnesses the power of grid computing, manages mixed environments, and provides customers with storage virtualization.
"The most immediately valuable features will be micro-partitioning, which lets users split up every CPU into up to ten pieces, each of which looks to applications like a completely independent server, yet is completely fault-isolated and security-protected from the others, and IBM Director Multiplatform - a common console across IBM's different system kinds, and potentially across other server manufacturers' products as well," said Jonathan Eunice, a Principal Analyst with Illuminata.
IBM claims that its years of experience in the mainframe space have allowed it to design "micro-partitioning" technology. IBM provides the example of a four processor system becoming a 40-way system running multiple operating systems. Micro-partitioning also includes virtual networking, memory and LAN capabilities.
Rival Sun Microsystems recently updated its Solaris operating system to divvy up processing power in much the same way, while also offering storage virtualization.
IBM Director Multiplatform is the company's answer to the staggering cost of pooling together and administering all of the pieces of an IT infrastructure. It works by serving as a single point of control for IBM and third party systems, clusters and grids. As a result, IBM claims that customers will no longer need to train personnel to be knowledgeable of distinct systems and will realize lower cost of ownership.
Workload management and provisioning tools powered by IBM Tivoli Provisioning Manager are also included. Big Blue says that its provisioning technology can be used to add new servers to meet spikes in demand in minutes – not days.
With its Virtualization Engine, IBM is supporting distributed systems based upon on Open Grid Services Architecture and WebSphere technology. Grids form the basis of a new type of supercomputer by pooling together physically separate machines to balance and solve complicated tasks.
While initially the gilded plaything of researchers, grid computing is being harnessed in finance, engineering, and even the battle against cancer. Grid computing can be sized to the task at hand: Campus Grids are local, while Global Grids can span countries.
Business information can also span multiple independent storage devices. According to IBM, "TotalStorage Open Software virtualizes and centralizes storage across heterogeneous storage devices to help clients optimize utilization, improve application availability, and increase administrator productivity." This component of the Virtualization Engine utilizes IBM's SAN Volume Controller and the SAN File System.
"Virtualization Engine is a package with a lot of different parts, but the basic idea is to bundle together a good number of virtualization components from across the eServer lineup and ship them with every server," Illuminata's Eunice told BetaNews.
You have to wonder who scooped who. Steve-o's executive letter says MSFT is commited to a platform with low administration, where IT dollars can be spent on advancing the office technolodies, to that end David Worthington of Beta News is reporting on IBM's 'Virtualization Engine' is stepping up beyond what MSFT's Virtual Server is offering.
MSFT is doing a fantastic job allowing you to minimize downtime, and creating a platform that is hardware independent, but IBM is upping the ante with its ability to virualize machines accross a grid. If you have lots of little server this means you can have 20 servers running on 5 cpu's, or conversely if your servers are monoliths you can have 3 servers running accross 10 processors. Where the win really happens is if you have servers that have varried usage.
Often I have CPU intensive work that takes days to run, but by virtualizing the tasks I can have a process that would normally take 24 hours once a week happen in less time using the spare cycles of the grid, and have the horsepower that I had been using only 15% of the time be available to the rest of my farm.
Combine this the improved uptime that can be gained by automatic fall over when hardware fails and you will soon be looking at a world where one guy manages a grid that sits in a rack in a basement, and he does this from his home, only making trips to the basement when hardware fails at which point he slide the server out of the rack, plugs a replacement in to the grid and life keeps going.
MSFT is playing catch up as Linux is already set up to work this way, where as Windows Server is not designed in a way that lends itself to parallell processing. And as Linux you can recompile applications to quickly enable existing products to be more compatible with the virtualization, Windows Applications from third parties can't be easily modified. This ties MSFT to maintaining 100% compatibility with the way theads and memory handling is presented to applications.
Where MSFT is leading is that they have the better addministration tools, allowing you to manage those server more easily. Just putting them on a hardware independent platform isn't enough. You still have to manage the security patches, administrate users, and do ongoing maintenance to the Exchange, SQL, and data stores.
Building easy-to-use, scalable solutions that cover every aspect of the management experience, with real-time feedback on system performance and a high level of automation
And this is where MSFT really is winning MOM rocks, and when you combine it with Virtual Server and Virtual PC, you can drastically lower your Personell requirements.
IBM's solution lets you run 10 machines as one server which can drastically lower your software requirements but at the expense of hardware.
This makes sense since MSFT is in the Software Business and IBM is in the hardware busines, that each tries to lower the cost of the bits they don't sell.
Brandon Wirtz
www.griffin-digital.com
Score: 0
|Just curious if anyone knew just what tasks this would allow performance betterment in. By performance, do they mean more processes handled, or do they mean the ability to do more cheaper like buying one machine and having it appear as 10 others to simplify email systems, web hosting, database hosting, etc. just curious.
Score: 0
|More load on reduced hardware is the focus. It works as long as colocation doesn't force hardware upgrades to the point where splitting them back out again (on middle-tier hardware) would be cheaper. It's a diminishing-returns scenario. You can colocate some server roles with good results (with enough disk and memory resources), especially with a Linux host OS because it's also cheap and I/O doesn't suffer too bad. But trying to host web, file/print, email and database services on one box is stupid unless you invest heavily in partitioned or attached drive arrays and tons-o-ram. It is excellent for labs and for software development testing though.
Score: 0
|Did anyone notice the conspicuous lack of any mention of VMware? Geez. Just who owns the VAST majority of that market segment anyway? Hmmmm....
Score: 0
|