100-Gbps TOE, in a near future…

According to ByteandSwitch there is a new startup called Grid-X (with main focus on Grid Computing technologies), that promises to develop (in cooperation with U.S. government) a 100-Gbps TCP/IP offload engine (TOE).

TOE is a technology used in high-speed Ethernet systems for the purpose of optimizing throughput. Briefly, it consists of a network adapter that performs some or all of the TCP/IP processing on an Ethernet adapter. This is achieved by removing the burden of I/O processing of data flowing through the network from the microprocessor and I/O subsystem to dedicated components on the NIC (network interface card) or host bus adapter. These dedicated components can be either specific processors or ASIC (Application-Specific Integrated Circuit).

As such, TOE solves the bottleneck that is currently created when the CPU reaches it’s maximum load and turns itself incapable of processing the Ethernet frames at higher rates, leaving also no processing resources left for applications to run on the system.

Currently there are 10-Gpbs TOE implementations, but a 100-Gpbs TOE could have a much bigger impact on data centers and HPC facilities, as it may allow Ethernet to substitute fiber links (Ehternet would be much cheaper than fiber-optic links, mainly due to the optic connectors price).

Anyway, as stated in ByteandSwitch…

Don’t expect to see many more details anytime soon, though. “Grid-X will be in stealth for a good year,” explains the source, who asked not to be named, adding that the firm’s TOE technology will be generally available sometime in late 2008.

so, it seems we will have to wait and see…

Google fight over “Grid vs Distributed Computing”

Hi there!

inspired by j.pereira post over Google Fight, I also decided to take a chance and do a fight based on the previous post… check it out on Grid Computing vs Distributed Computing!

What about Grid Computing?

So… what do you know about Grid Computing? isn’t it just another form of Distributed Computing? well, not quite…

The concept of Grid Computing goes further than that, as it reassembles not only the capacity to share traditional unused resources like computational power (CPU cycles from loosely coupled computers), but also the ability to manage and share de-centralized data storage (disk storage resources), control user access to resources in order to comply with desired SLAs and providing a complete security infrastructure to support the communication between the distinct software components of the Grid. Preferentially, all of this using Open Standards.

The definition of Grid was initially proposed by Ian Foster and Carl Kesselman, back in 1998 (“The Grid. Blueprint for a new computing infrastructure”). More recently Ian Foster presented an updated article (“What is the Grid? A Three Point Checklist“) were a kind of Grid checklist was presented, stating that a Grid is a system that:

1. “coordinates resources that are not subject to centralized control“
2. “using standard, open, general-purpose protocols and interfaces“
3. “… to deliver nontrivial qualities of service.“

As such, the concept of Grid Computing also introduces a kind of virtual layer for resource sharing, decoupling the applications and workload management functions from the Grid Infrastructure hardware.

I’m not going into deeper explanation of what a Grid is (because you guys would be bored!), for that please check Wikipedia here.

And what about current implementations of Grid Computing systems? is there any? you bet! mainly, Grids are used for Scientific purposes, and there are some big deployments like the Open Science Grid, EGEE (Enabling Grids for E-sciencE) and probably the most known DataGrid Project and LHC Computing Project (LCG) at CERN.

All of the above mentioned implementations use the Grid technology as a way to process huge amounts of data from several scientific projects. Just to have an idea, when the LHC (Large Hadron Collider) at CERN starts working (expected to be during this year) it will generated about 10PetaBytes of data per year that needs to be processed!!!

The technology behind a Grid doesn’t rely only on software components (like the Globus Toolkit, the most recognised open source toolkit to develop Computer Grids), but also strongly depends on the networks that make possible to share such huge amount of data across the whole planet. Trying to implement such a global grid some years back, was just impossible due to bandwidth constraints!

And as any other technology that starts in the academic and scientific area, there must be some big guys also interested on this issue, right? right… the most active Industry players have been IBM, Sun Microsystems and Oracle.

…this post is getting too big, I will just leave some links below in case you want to digg a little bit more on Grid Computing:

• Open Grid Forum – http://www.ogf.org
• The Globus Project – http://www.globus.org (check also the publications section)
• CERN Grid Caffe – http://gridcafe.web.cern.ch/gridcafe
• IBM developerWorks > Grid Computing http://www-128.ibm.com/developerworks/grid
• Oracle Grid Technology Center http://www.oracle.com/technology/tech/grid/index.html

Well, for know this is it! just a small insight on Grid Computing… but I expect to continue later on with this Grid stuff, there is still a lot to talk about, especially on what concerns the exploitation of Grid Computing to make Business! better said, how to make money from Grids… where can we use them?… any hint?

cheers!