Home · About · A-Z Index · Search · Contacts · Press · Register · Login

Plenary - Boundaryless Information Flow:
Grid Computing

Day 1: Monday 21st July 2003

What is Grid? - We examine the emergent technology, consider the various interpretations of Grid, and the technological challenges that Grid seeks to resolve.

Introduction

Allen Brown, President and CEO of The Open Group

Allen began the Conference by welcoming the delegates to Boston. He extended a particular welcome to the large number of non-member attendees to this well-attended Conference, and invited all to take an active part in the meetings, and to discuss things of mutual interest.

Grid is a fascinating subject and in this Plenary we aim to cover a full spectrum. We will start off by looking at what Grid is, and then discuss if it is "ready" for use. Then we will look at the vendors' viewpoints - we have Sun, HP, IBM, and Oracle - and then move on to look at requirements and security issues. Then tomorrow we will take in more customer viewpoints, then look at the management, legal, and policy issues involved, follow this with how the Grid is being realized by the Global Grid Forum, GRIDS and Globus, and close the Plenary with a view on Boundaryless Information Flow and the Grid.

Allen introduced the keynote speaker, one of the leading forefathers of the Grid, Professor Ian Foster, who is Associate Director of Distributed Computing at Argonne National Laboratories and Professor in the University of Chicago, which together form the home of the Globus toolkit which is the de facto standard for open systems Grid computing. He is also the author of a book "The Grid Blueprint for a New Computing Infrastructure".

Keynote Address
Professor Ian Foster Associate Division Director & Senior Scientist, Argonne National Laboratory

Ian opened by addressing three immediate questions:

• What is the Grid?
• Where is it today?
• Where is it going and why should you care?

He presented six statements on what the Grid is, and agreed it is all of them. The origins of Grid are in eScience, and its rapid development has been fuelled by the open source Globus toolkit. At present we are experiencing rapidly growing commercial adoption focused on intra-enterprise resource sharing, based on the Open Grid Services Architecture (OGSA). The future of Grid is seen as the key enabler of new applications and industries based on resource virtualization and distributed service integration. The reason we should care now is that Grids address real resource-constrained computational pain points now; Grids are a disruptive technology that are ushering in a virtualized, collaborative, and distributed world of IT; and an open Grid is to the advantage of all IT users so we should all encourage adoption of the OGSA approach and refuse proprietary solutions.

Ian went into some detail on the pre-Internet and post-Internet origins of Grid computing, and how these relate to today's business drivers which are demanding large computational resources both for heavy computational resources and for instant on-demand responses during peak load-times. He developed this in a slide representing today's Enterprise Computing Environment.

Common eScience/eBusiness requirements are to have a virtual computing system that is dynamic, multi-faceted, spans institutions and industries, and can be configured to meet instantaneous needs, can handle demanding workloads, and can dynamically link resources and services from collaborators and customers and eUtility providers, who then become members of an evolving virtual organization.

So Grid technologies address these requirements directly, both as an infrastructure for establishing, managing, and evolving multi-organizational federations, and as mechanisms for creating and managing workflow within such federations. Looking at several patterns of use:

• Resource/service visualization - encompassing discovery of resources/services with the required capabilities and availability
• Resource/service integration - matching the requirement to availability of the right resources at the right times to fit an acceptable quality of service.  Ian described an example of a Grid Computing Portal in the University of Texas at Austin; and also Platform Symphony for real-time online processing, which uses virtualization at application and infrastructure levels to make an extremely powerful system.
• How Grids involve multiple independent organizations, and the consequent need for negotiating acceptable cross-institutional organizational parameters. He used aviation safety as an example of a cross-institutional approach in NASA aviation system behavior models.

A possible likely end-state of where we're going is decoupling of provision of computational services and consumption of those services. The economics of computing will, as a result, change rapidly, as new devices become available to access remote services, and provider services also rapidly increase to meet escalating new requirements and high availability demands.

In this scenario, looking back at the definitions for "what is a Grid?", there are three criteria that distinguish and characterize it:

• It coordinates distributed resources.
• It uses standard, open, general-purpose protocols and interfaces ...
• ... to deliver non-trivial qualities of service.

A Grid is not a cluster, a network-attached storage device, a scientific instrument, a network, etc. Each is an important component of a Grid, but by itself does not constitute a Grid.

In the Grid world, there are many success stories. The Globus Toolkit has enabled much rapid development and is a de facto standard for major protocols and services. The Global Grid Forum is pulling together the community and standards, and there is an emerging Open Grid Services Architecture (OGSA).

The OGSA was needed to provide a protocol suite that enabled a common base and exposed key capabilities, to have a service orientation to virtualize and unify resources/services/information, and to embrace key Wen services standards and leverage existing commercial solutions. The result is a foundation set of standard interfaces and behaviors for distributed system management - the Grid Service. Ian characterized this by describing OGSA as equivalent to Web services on steroids! - extending them to work well in a Web services environment.

The OGSA has led naturally to a complimentary Open Grid Services Infrastructure (OGSI) which defines core interfaces and behavior for manageable services, introducing the notion of discovery, creation, and closing of services. Associated standardization work is now underway with a soft lifetime for those services.

Hurdles that Ian sees for Web services and OGSA to overcome include:

• An industry and customer focus is needed to pass a maturity point before the next silver bullet that will represent the next jump-start for Grid computing.
• Standardization is extremely important to ensure portability of applications and interoperability of solutions.
• Unfriendly licenses - Ian asserted that inappropriate greed over IPR in the Grid environment can set the Grid computing movement back to where it will take a very long time to realize its true commercial potential.
• We need to learn from previous efforts, rather than reinvent some existing wheels.

However, Ian remains highly optimistic that Grids will succeed, because:

• Web services have lots of standards and interoperability issues and too many options - there is too little time for these to be resolved before they are overtaken by Grid developments.
• Grid requirements are a superset of Web services requirements - it a natural progression to move up from WS to OGSA.
• Globus already provides the basic working open source toolkit, which is being deployed in a growing, global, demanding user community. Vendors will use it - experience shows that nothing "sells" better than working free code.

We have to take care to manage the Grid silver bullet. Ian drew parallels with DCE, and with CORBA, but predicted that Web services will be succeeded by Grids through the success of OGSA plus Web services plus the Globus Toolkit.

Ian closed by doing a recap on what the Grid is, its past and present and future promise, and why we should all care about it as the next big evolution in computing. To summarize, Ian advised us to look beyond the Grid hype, to compare the added value of OGSA with Web services so as to realize how we have to move on from WS, and to look into the sophisticated plumbing and services that Grid offers. Vendors are commercializing "the Grid" now, Ian claimed, and he recommended all present to get involved with the Global Grid Forum to ensure progress is maintained along the right lines.

He ended with a reminder of GlobusWORLD 2004 (www.globusworld.org , 20-23 Jan 2004), plus the following URLs as useful references:

• The Globus Project - www.globus.org
• Global Grid Forum - www.ggf.org
• Background information - www.mcs.anl.gov/~foster

Top

Grid Computing - Getting Down to Business
Thomas Hawk General Manager, Grid Computing, IBM

Tom began his presentation by saying how good it is to follow an introduction to Grid by Ian Foster - it enables him to build directly on the understandings that Ian has established and go straight into how to apply Grid to ensure a sound business value.

Key messages are Grid is about business value, Grid is creating value for commercial enterprises today, open standards are critically important, and Grid is the first step towards on-demand computing.

Grid computing enables IT and business value. It improves asset utilization, integrates heterogeneous resources, enables data access and integration and collaboration, it strengthens redundancy and resilience, and it is quick and responsive. All this represents greatly improved IT value - the ability to do distributed computing over a network that can extend globally, using open standards to enable heterogeneous operations.

Tom explained how IBM sees Grid evolution as likely to follow a roadmap of increasingly powerful stages:

• Base grid
• Data virtualization
• Workload management
• Task scheduling
• Automated provisioning
• Billing and metering
• Transaction management

This will deliver to enterprise customers what is effectively eBusiness on demand - an enterprise whose business processes - integrated end-to-end across the company and with key partners, suppliers, and customers - can respond with speed to any customer demand, or market opportunity, or external threat. Tom said that IBM does see how to progress along this roadmap in ways that will not require any major change in existing technologies.

Tom illustrated his vision with an example of how a financial market would benefit from a grid computing solution, including how the customer would benefit. The business value of Grid computing can be measured in many ways but in his estimation they boil down to improving operating efficiency and return on investment, reducing capital expenses, accelerating business processes, enhancing employee productivity, and being easily and quickly adaptable to changing business requirements. He emphasized that OGSA is the rallying critical step in calling out the open APIs needed to enable open exploitation to take place. Competing companies want to succeed against competition - the buying community are going to remain the drivers of the market and power over how Grid progresses. There are huge productivity gains - imagine the possibilities and what it would mean if your business could:

• Analyze the value or risk of an investment portfolio in minutes rather than hours
• Significantly accelerate the drug discovery process
• Scale their business to meet cyclical demands – while cutting expenditures
• Reduce the design time of your products, while reducing instances of defects
• Unite their research teams with others around the world to take advantage of the most up-to-date learnings

Tom next looked at what he considers are the key Grid market focus areas - R&D, engineering and design, business analysis, enterprise optimization, and government development - and showed where IBM is active in responding to each one. He went on to give examples of successful Grid deployments in which IBM are closely involved - in Aventis, Royal Dutch Shell, IBM chip design and test, Charles Schwab, RBC Royal Bank, Kansai Electric Power, TIGER project in Taiwan. These demonstrate IBM's commitment and focus on making Grid computing a success.

Tom explained how he recommends getting started:

• Get educated
• Determine the value of Grid to your organization
• Identify the right Grid offering(s) for your business
• Develop and prioritize a list of Grid pilot implementations
• Architect and implement Grid solution(s)

He appreciates there are many ways to become engaged in the Grid movement, and encouraged everyone to decide which way is best for them, but to ensure they do make Grid a part of their IT business competitive strategy.

Tom closed by showing IBM's picture for a top-level architectural framework for Grid, with OGSA at its core, then mapping this onto the various operating areas within IBM that deliver solutions in each part of the framework.

Q: Andras Szakal (IBM) - Noting that middleware is not in place, why is this when IBM is essentially a middleware company?

A: Context is Ian Foster's comments on middleware not creating money.

Q: Rob Mitchell (Computerworld) - Putting a timeline on Tom's roadmap, how do you see it mapping out?

A: Tom couldn't put a timeline on it because there are so many variables and unknowns. However, he sees some very positive steps underway already so indications are encouraging.

Q: Allen Brown - Open Group members believe in collaborating to produce open standards - how should we engage?

A: Tom - Do it through the GGF - IBM is fully supportive of this approach.

Q: Col Logan - context?

A: Tom - standards around integration of information are rapidly evolving. Also there is much activity in the management and federated data space, including on Data Access Integration WG within GGF.

Q: Ray Chang - What's the status of tools to enable IT engineers to get their applications running?

A: Tom - Tools are critically important - they are still at relative infancy at present, but the long-term aim is to deliver out-of-the-box applications. This issue is on the list of important things to do, and it needs to be prioritized.

Top

Is Grid Fit for Commercial Use?

Grid Computing, The Industry View
Wolfgang Gentzsch Director Grid Computing, Sun Microsystems Inc.

Wolfgang said he wants to fill in from Ian Foster's and Tom Hawk's presentations, to give Sun's industry view. Sun supports the standards work that is underway. Sun wants to use Grid as the core layer to support computing as a utility - just like we consider electricity, water, telephony, etc. as utilities.

In this sense, Wolfgang said the characteristics of a computing utility are that it allows you to get what you want when you need it, you don't care where is comes from or how, and you pay as you go for what you used. This is Sun's vision for Grid.

Wolfgang listed key benefits of Grid as the next generation enterprise architecture:

• Access - seamless, transparent, remote, secure, wireless access to many resources
• Virtualization - there is no such thing as a Sun grid, or IBM Grid, HP Grid, etc. Access is to computer and data services
• On-demand - you get resources you need, when you need them
• Sharing - enable collaboration over the network
• Failover - migrate and restart applications automatically
• Heterogeneity - platforms, operating systems, devices, software
• Utilization - increase resource utilization from 20% to 80+%

He compared High Performance Technical Computing (HPTC) Grids versus Commercial Grids - suggesting that the difference in requirements are significant, and noting that the HPTC side includes innovators and early adopters so is necessarily harder to progress. Over time his vision is that, thanks to OGSA/OGSI, we will see more convergence on IT services. Sun's evolutionary Grid strategy is to move from cluster to enterprise, to Global Grids. The customer demand to improve return on investment (ROI) is high, and the rapid growth in the 1990s in client-server departmental computing has provided much spare networked computing capacity which is now available for Grid computing if the business decision is there to make it available as a Grid resource.

So how is Sun doing this? Wolfgang said that as a system and infrastructure company, it makes sense for Sun to build the hardware and middleware to enable customers to run and utilize the Grid efficiently. This gives the configuration that is Sun's Grid Services Environment. The core of this is their Grid Engine family, developed with the Globus Toolkit. They have also extended this into an enterprise edition which adds alignment with business goals via policies. The gobal layer includes Avaki. The e-Science Grid is also important, having enabled major academic collaborations for Sun, especially in UK. In a further slide, Wolfgang listed many other organizations that Sun are involved with in Grid computing projects.

As for any infrastructure, applications for the Grid are key to its deployment and adoption, and these cover single CPU jobs, array jobs, massively parallel jobs, and parallel jobs. Wolfgang's slide details the nature of these different kinds of application.

The Grid increases complexity and we need to get rid of this complexity; virtualization helps here by enabling management of resources rather than managing each server, and servers sharing resources. Wolfgang illustrated this in the context of the Solaris F15K box and the new N1 network operating system for the datacenter.  The N1 effect on efficiency has been enormous, producing radical improvements in performance, reduced costs, and uptime.

Top

What is the Grid and what will it do?
Greg Astfalk Chief Scientist, Hewlett-Packard

Greg noted that Ian Foster, Tom Hawk, and Wolfgang Gentzsch have not colluded on their slides but they have many similar messages.

The Grid has tremendous potential in the IT community at large, from the consumer space to the enterprise. We are not yet ready to declare success - pervasive mainstream use of Grid will take some time, and to make it a success an open community effort is essential.

HP views Grids as implementations that support all the tasks associated with Grid services. The Grid can allow for secure sharing of and collaboration with Grid services between mutually consenting organizations in a virtualized computing environment. Greg believes Grid will do for computing what the Web did for information.

Greg sees Grid "mega-trends" as abundant bandwidth; global reach; high complexity of management of accumulated IT resources; less-and-less centralized, dispersed ownership of the content of science, commerce and consumers; and potential for shifting of computation requirements around the globe so that the computers of those who sleep can be used by those who are awake and working in other timezones. Greg agreed with previous speakers that Grid has its roots in scientific computing, and it will continue to be relevant and important there. However, the greater importance of Grid is in the commercial space. OGSI - where Grid meets Web services - is where he believes the commercial importance lies.

Grid marketing projections include the Gartner Group which urges caution against investment in Grid until clear tangible business value materializes. A few do make projections and loosely agree that Grid spending will grow from $250 million in 2003 to $4 billion in 2008. It is unclear how these numbers break down - hardware, software, services, etc. Predictions are limited to a few narrow markets - life sciences, energy, manufacturing, public sector, etc.

Greg considered the key requirements for a successful commercial Grid - bearing in mind that commercial business decisions are made by CIO business managers not IT experts - include management, trust and security, heterogeneity, open standards at the Grid computing levels, and robustness. Without these requirements being met, the commercial deployment side will flounder.

There are also social engineering issues - loss of control of resources, loss of IT funding, resource sharing, server hugging, and change that humankind has to accommodate to become comfortable with the new Grid computing approach. CIO demands for a new IT enterprise architecture include a platform for managing change that links business and IT, based on standards in their entire IT environment, which eliminates vertical islands of automation, embraces heterogeneity and legacy IT environments, uses automation to scale and reduce complexity, virtualizes all IT assets, and helps convert fixed costs to variable costs.

HP development in Grid includes developing technologies in several areas:

• InteractiveGrid
• SmartFrog (Smart Framework for Object Groups)
• Utility Data Center (UDC) + Grid:
  HP have also developed a Utility Data Center (UDC) - a wire-once, programmatically reconfigurable, virtualized data center with fine-grained allocation, security, and control of every resource - and they have developed a Globus/UDC interface. The two provide a powerful combination.
• Fine-grained Grid
• Appliance aggregation
• IPv6
• Grid and Grid Services Management
• An Intra-HP Grid involving about 70 nodes, for test and development
• A Grid-lite which would be more suited to the consumer services market

HP has also been working on integrating their Adaptive Enterprise (AE). Their approach has been to conceptualize their AE as having three layers and an adjacent tower. The layers are business practices, Grid services providing “connective tissue” of applications and business practices to the infrastructure, and an infrastructure. The tower is management of the infrastructure, the Grid and Web services, and the applications and business practices.

Greg added that HP has developed a Web Services Management Framework (WSMF) to manage Web services, Resources, Grids, Grid services, Applications, and business practices. These are all compliant with standards efforts (OGSI, OASIS, etc.), and interoperable with CIM, SNMP, JMX, etc. Greg anticipates that HP will likely contribute this to GGF.

In conclusion, Greg emphasized that Grid is a strategic technology for many parts of HP, and that management of the Grid is crucial to its success. In addition, their focus is on pursuing robust Grids for returning commercial value, pushing and adopting fully open standards, and developing management tools for Grids and Grid services.

Top

Fujitsu's Challenges in GRID Computing
Kimio Miyazawa General Manager Information Technology Core Laboratories, Fujitsu Laboratories

Miyazawa-san said Fujitsu has much experience in development of Grid computing - he listed several projects as examples - ITBL, Super-Sinet, VizGrid, UNICORE, NAREGI. They see the main challenges as how to integrate Grid with Web services and Organic Computing, and just like previous speakers - how to establish a success story for Grid not only in the scientific domain but also in real business.

Fujitsu's strategy to realize the new Grid paradigm is code-named Triole. In this approach, they have established relationships between autonomy, virtualization, and integration. They believe Web services can be powered by Grid. They have a strategy focussing on Grid, in which virtualization of IT resources, applications, and databases can provide the platform. The key is providing applications. He took manufacturing, financial, and distribution as examples of industries that are demanding greater computational power, which can only be delivered through Grid solutions.

There is vast spare computing resource in the world. Fujitsu have mapped these into its GRIP (Grid Integration of Platform) framework. Characteristics of GRIP are that it monitors the state of all resources, matches the most suitable resource for each job, provides Organic Job Control, and provides real-time feedback of computing results.

Field trials of GRIP in the CAD application environment produced performance improvements of 3x in turnaround time and 4x in manpower. These are stunning results for the potential business efficiencies that Grid computing can deliver to a business.

Miyazawa-san went on to explain Fujitsu's interest in the Data Grid. The objectives are virtualization of heterogeneous and distributed database systems, and realization of data migration. He outlined the basic architecture  involved, the organic storage system for data migration, and the access grid for virtualization of IT resources for users and for process integration. He also described how they are involved in a collaborative research environment for experiments and simulations.

He acknowledged that utility computing technology is getting significant attention as the way to go to meet demands from customers, and showed examples - a framework for accounting, and a linkage between local and central IDC - demonstrating Fujitsu's involvement in this approach.

Miyazawa-san concluded that the way forward is Web services powered by Grid computing plus Fujitsu's organic computing manager TRIOLE.

Top

Oracle and the Grid
Brajesh Goyal Principal Product Manager of Distributed Databases, Oracle Corporation

Brajesh noted The Economist (June 21, 2001) said the best thing about Grid is that it is unstoppable. This is coming true now. He agreed with earlier definitions of what Grid is - computing as a utility, as a client-side view, as a server-side view, and as virtualization of provisioning of resources across sources of data (information) within a data center. The benefits are better information faster, faster response to changing business priorities, and reducing IT costs.

Brajesh asserted that the biggest IT technology trends at present are faster deployment by using blade servers (Brajesh asserted that all the major vendors are offering them now), rapid growth in deployment of Linux (Linux and blades naturally complement each other), and storage (NAS, SAN, and IB provide storage access from any blade). Economic reality calls for Grid computing, and the momentum is building up. The next phase of the Internet will have to be Grid-powered computing with Web presentation. For these reasons, Grid computing is unstoppable as the next big thing in IT.

Oracle offers its Oracle9i as the database of choice for the Grid. Brajesh listed key differentiators that make their database systems premier candidates for the database of choice in Grid environments. He asserted that Oracle9i:

• Uses blade servers to ensure best performance, flexibility, expansion capabilities, and robustness at optimum cost
• It can unplug and plug database data files
• Its streams feature enables sharing of data flexibly
• It is a distributed database which will work with other databases
• It offers superior and well integrated operational characteristics
• It is portable and offers easy migration to use in the Grid
• Packages all that is required in a single fully supported product package

Grid computing is core value for Oracle.

Top

The Development of Grid - Applications, Standards, and Security

Ian commented that we have heard many definitions for Grid but what is more important is making it work.

What's holding Grid back?

• Non-technical barriers. Platform has surveyed large consumers and found that organizational politics are the greatest barrier - and these break down to loss of control, risk associated with enterprise-wide deployment, loss/reduction of budget, reduced priority of projects, lack of data security between departments, and fear of external leaks.
• Lack of standards. Customers demand standards-based solutions in products they buy. Platform is a proprietary applications solutions supplier, but needs standards on which to build applications that give added value but which run on a standard framework. OGSA is the most significant Grid standard, and the Globus project and GGF have moved this forward very successfully so far. However, OGSA is an incomplete standard, and OGSI is seen as the way forward to ensure applications compliance with the Globus GT3 Toolkit.
  The Community Scheduler Framework (CSF) is an open source meta-scheduler which is compatible with OGSI and provides basic protocols to help resources work together in a heterogeneous environment. It is a development from Globus, IBM, and Platform. Platform is contributing its first meta-scheduler software to the Globus Toolkit 3 in Fall 2003.
• Lack of Grid-enabled applications.

Platform believes that a mix of open source and proprietary software is important to fulfill requirements for Grid computing. Platform is an NPi founder, and NPi merged with GGF in April 2002. Platform is committed to open standards, and has demonstrated this commitment through years of experience in the open source area.

Ian said Platform is in the business of accelerating intelligence, delivering intelligent, practical, enterprise grid software and services to plan, build, run, and manage Grids. They have a practical approach driven by business requirements. They adopt an end-to-end enterprise view to optimize existing resources, a deep understanding of business processes and priorities, and unmatched global service and support, all proven by over 11 years of service and satisfaction to over 1,600 customers worldwide, plus industry collaborations with several major vendors including IBM, HP, Dell, SGI, and NEC.

As tangible examples, Ian mentioned ASCI (Accelerated Strategic Computing Initiative Grid) which connects scientists securely in the World's largest Grid, TACC (Texas Advanced Computing Center) - a "grid-of-grids" for scientific research, and a collaboration with Pfizer on accelerating drug discovery.

Top

GES and its Requirements for Grid
Rob Walker Project Manager, Department of Defense’s Global Information Grid (GIG) Enterprise Services (GES) initiative, DISA

Rob started with a review of where we are today, having moved from systems in 1970s to System of Systems in the 1980s, to Family of Systems in the 1990s, and now to Net-centric Global Information Grid (GIG) in the 2000s. We now face the Net-centric challenge.

Part of this challenge is:

• To change the way we think and prosecute a different approach to how we operate and govern
• Need robust communications
• Need services to exploit the network and support applications

The GIG Enterprise Services scope is best represented as joining two enterprise services hemispheres:

• Net-centric enterprise services (NCES program)
• Domain and COI enterprise services

and the problem statement is how to have the two collaborate and share in the way we manage IT services. Also how to scale the associated Service Level Agreements (SLAs) to make them appropriately deliverable down the defense consumer chain. There are also significant social engineering issues in deployment, etc. Despite these barriers, they see Grid computing as a realistic solution.

Some of the Grid requirements include services, security mechanisms, and tools to Grid-enable applications.

Rob then gave the DoD view of how Grid applies to the required transformation - to transform defense intranets into powerful self-managing virtual computers, to fit with the DoD vision for a GIG, to integrate with the DoD's cast range of heterogeneous systems and the huge investment base this represents, and to fit with the DoD vision for GIG Enterprise Services (GES) which will provide ubiquitous service-oriented access to DoD information.

In the context of the GIG ES, the focal points are Discovery, Storage, Application-hosting, Information Assurance and Security, and Enterprise Services Management. They have many issues needing resolution, among which are:

• There are many different definitions -  the real meanings are still in flux
• How to keep control and security in the abstracted hardware layer
• There are lots of competing players
• How to decide and prioritize Grid requirements with respect to networks as computers

However, there is no doubt that Grid computing is the next generation environment to address largescale processing challenges. This fits with the GIG ES requirements, but the computing resource management side of it is very immature at present and is the key to gaining confidence that it will work securely and resiliently without loss of necessary control.

Top

Grid - The Standards Issues
Professor Ian Foster  Associate Division Director & Senior Scientist, Argonne National Laboratory

Ian gave an overview of the relevant standards activities. First he restated why we all appreciate that standards are important. We are making progress in establishing the key standards. The Global Grid Forum (GGF) is working on OGSA and OGSI. The W3C is working on Web services (WDSL, SOAP), OASIS is working on Web services (SAML, XACML), IETF on Internet protocols and security, Liberty Alliance on Identity Federation, DMTF on Management.

In the Grid world, the Open Grid Service Infrastructure (OGSI) will make an application portable. It will link a Grid Service Handle with a Grid Service Reference. Also it will provide clients with information on the Grid service required from Service Data Elements. Additionally it will provide lifetime management for the service.

Ian explained that the OGSA working group is coordinating GGF standardization work to include the OGSI work. Other GGF working groups are addressing the following areas:

• OGSA security
• CIM-based data models
• Data access and integration services
• Agreement negotiation
• Resource usage records and protocols
• Common Management Model
• Base Management Port Types
• Lifecycle Management Port Types

and there are many more working groups needed to address the whole problem space.

On security, Ian said they are building on IETF security standards - PKI, Kerberos - and use X.509 certificates in the Globus Toolkit. Web services security is more complex, and needs to be reconsidered and perhaps extended for OGSA - there are still many holes to fill. The GGF security working group is coordinating security standards for OGSA applicability.

On Internet issues, they are working with W3C on WSDL and SOAP, and have OGSI standards authors working in the WSDL v1.2 team to ensure alignment with OGSI.

The IBM/Microsoft WS Security Architecture was announced in 2002, and is being revised with an eye to being integrated into a practical implementable solution - much more work is needed here.

IPR issues are a major concern at the present time, and Ian expressed fears that the legal issues involved here could be exploited by short-sighted players to gain short-term benefits that will undermine all the technical progress to bring the huge performance benefits to the enterprise marketplace.  Ubiquitous adoption is likely only to happen if IP is licensed royalty-free. The OGSI authors have made a commitment to this royalty-free objective.

The Globus Toolkit continues to adopt open standards as they emerge.

Summarizing - Ian said standards are critical to the success of Grids, it is a complex space with many stakeholders, GGF is defining key Grid standards (OGSA, OGSI, data, manageability), collaboration with other standards bodies in this space (W3C, OASIS) is ensuring good alignment on applicable standards, though uncertainty on security standards remains a concern. Open source software remains vital to encouraging development of applications that will make Grid useful to business.

Top

Allen Brown introduced the panelists:

Ian Foster, Wolfgang Gentzsch, Greg Astfalk, Brajesh Goyal, Ian Baird, and Rob Walker

Q: Chris Greenslade - In the slide on a silver bullet for Grid - if you had the quick chance to explain Grid computing, what three items would you mention?

A: WG - Start with the current situation - low utilization and potential for easy access and vastly improved productivity.

IB - Do you want to be able to do more in a shorter time and reduce cost - these are the realities when we install a Grid. Also set up a pilot departmental Grid to demonstrate the dramatic gains in productivity before extending it to the whole enterprise. Once these huge gains in productivity are seen by a CIO/CEO these decision-makers invariably overrule most other objections.

RW - As a DoD person he would probably be talking to a military leader - so would put it in terms of how to more effectively provide more support and save lives using much less resources.

BG - Better utilization at much lower cost.

Q: Is OGSA something to Grid-enable my applications - if not, what is?

IB - OGSA is not; a toolkit is, and this is under development at present.
IF - There are tools to help this now. However, we will not see a single solution.

Q: Eliot Solomon - It seems Grid is all things for all people. Are we likely to see different Grids for different purposes?

GA - Sees at least 2 types of Grids, 1 a Grid-lite for consumers, and another for industrial-strength; both must be compatible.
RW - Still expect to have a mix of existing systems in the DoD with Grid-based applications running alongside existing applications, so he anticipates a steady migration from the $100billion existing investment to a lower level of investment.
BJ - Believes there will be a single Grid in the long-term
IB - Grid is a notion right now, not a deliverable, and it will take some years before the real Grid of the future takes on a stable form.

Q: Eliot Solomon - Is any attention being paid to service directories?

A: IF - There has been no systematic analysis to date, but this is also an issue on the list of items to prioritize.

Q: Bob Blakley - Several presentations dwelt on unhappiness with existing standards work, on IP terms, and of course availability of commercial applications for Grid - in an ideal world, what do you want to see here?

A: IF - Doing standards work in closed groups is damaging, so he would like to see more open collaboration on standards.
RW - From an implementation point of view, how do I make decisions to buy when there is a lack of implementation experience. Good things are coming out of standards work but it is not keeping pace. Early innovators are rolling out products but these are not for the commercial business customer.

Q: Andra Szakal - As a software architect he dislikes misnamed technology - so what are the panel's comments on the description that Grid is an application of autonomic application?

A: IB - Rationalize architecture and Grid computing.

A: IF - Whether it's a good name or not is not really material now - it's stuck and established. Things have moved on and they are now trying to find applications and interfaces that are scalable and that will meet the need. No one name will fill the overall need and objective.

Q: Silicon Graphics - No-one has talked about visualization across the Grid. If you did, what would it look like?

A: RW - In the military, visualization is an important aspect. However, he is not sure what Grid should do to provide this at this time.
WG - HPDC includes always, all the time, everywhere, and graphics focussed resource is a part of this concept.

Q: Grid products are not based on sound security - how does the panel see applications coming out in a sound foundation and secure environment?

A: IF - We are working in the GGF on a sound theoretical foundation, and aim to develop frameworks for these, but as yet we don't have a clear vision for what this will look like.
RW - NSA is taking a strong interest in this (MILS). However, we can secure applications on an Internet level for the present. There is always a decision to be made when we can roll it out.
IB - His notion to roll out enterprise Grid on a Department basis then grow its deployment is best here because it verifies what the real security needs for that business really are.

Q: Allen - What should anyone considering putting in a Grid list as the main issues to address?

A: IB - He has found that political issues are the most important concerns to be overcome - loss of control, loss of budgets, etc. The best solution to this has been his ability to set up and run a model to demonstrate how inefficient not using Grid can be, and also to show how rules and policy-based elements can ensure those with concerns do not have to lose control of their system. When they see how much productivity it produces the old arguments tend to disappear - the CIO/CEO overrules and reorganizes to remove the problem.
RW - Believes there is nothing more dangerous than a project manager with money - the initial need is to prove through case studies that deploying a Grid will produce significant savings. When you get buy-in from the top that it does produce huge efficiencies at reduced cost it becomes self-fulfilling.

Q: Tim Childers - With a pool of computing resources, does this not suggest interest in seamless fault tolerance for the Grid?

A: IB - Yes, and they call this business continuity. He has worked with clients who use under-used capacity in organizations to mirror systems.
GA - In Grid you don't know who steps up to provide the resource so this issue is moot.
BG - Oracle will take care of things for you.
RW - Agreed with IB that they too call it continuity of operations, so he would expect to extend this to their Grid systems.

Top

Securing the Grid
Dan Geer  Chief Technology Officer, @stake

Dan said what Grid computing needs is:

• Reliability - security is a subset of reliability - if it is insecure then it is unreliable. Security is a means, not an end. If you accept this, mine the wealth of existing literature to find out what you need.
• Independence of location - OK something has to move! The more things are in motion the more risk you have.
• Economy. Economics always wins in the end - it's why we're here. Costs have reduced as computing power has risen. 
• Measurability by design is the only answer.

Core security requirements are more of the same (but a lot more): integrity of hosts and results, verifiable metering, and confidentiality (of action as much as data). Research-grade problems do exist. Thinking in terms of protecting the players in the Grid - the Grid provider protects self from customer, and customer from self, but who protects the customer from the customer?

Considering tractability of the problem for applications and security, applications are where the action is now, and this is especially relevant for Grids. So if there are no new issues, then it is very important indeed, because what makes it harder is:

• Applications are federating - this creates complexity over knowing where the application actually is.
• Perimeter defense diseconomies - asking "are my applications secure?" brings into question if perimeter defenses are effective in a Grid environment. The answer is they are not.
• Data is everywhere, applications are getting more complex, and the data they use is proliferating hugely. The cost of storing data is insignificant.

We do need a common language because we need metrics. We don't have them right now, but we really do need them to be acceptable in business terms, because customers want to know answers to reasonable audit-type questions like:

• How secure are they?
• Is their security improving?
• Are they spending the right amount of money on it?
• How do they compare with other organizations?
• What risk transfer options do they have?

Answering a question, Dan agreed that we can't have metrics without content; we can't prove a negative, but it's pointless to try to do so - witness airport screening. However, you can try to make a realistic estimate of the level of effort to break a system, and you should always be thinking in terms of whether you do have enemies who are willing to put that level of effort in.

Dan asserted that security of applications is improving; he used statistical experience of surveying his own customers. However, the normalized spread amongst applications is getting broader, which probably means that the counterparty risk is rising. He uses numbers - metrics - because that is the only way to get objective comparability.

Dan believes that with all this increase in traffic and data, accountability should be a Grid design goal. This begs the question of detecting anomaly, which is compatible with dissolving of perimeters and with the assertion that selective data deletion is not worth the cost saving in releasing data storage because data storage is so cheap these days. This would mean that the access control model should be replaced by an accountability model. Dan submitted that the overall answer for security is accountability - this should be the design goal, because it is the only practical alternative.

This goal is also compatible with the consequence of accountability as the security mechanism, because huge computing power will be needed to do the necessary surveillance to detect anomalies, and Grid provides exactly that huge computing resource:

• Target of choice versus target of chance
• Traffic analysis
• Forensic quality data - cheap to retain
• Replication for reliability (hence security)

Top

Allen noted that we are continuing this discussion the next day. He invited all delegates to attend the Monday evening networking event, during which there will be an entertaining non-technical presentation from Ken Ashkenas, author of the book "The Boundaryless Organization - Breaking the Chains of Organizational Structure".

Top

Guest Speaker Reception

Beyond Technology: The Human Challenges of the Boundaryless Organization
Ron Ashkenas Managing Partner at Robert H. Schaffer & Associates

Ron noted that change is a business imperative for organizations who want to remain competitive and relevant to their evolving goals. It is also a management tool that is sometimes used unnecessarily. His experience of how Jack Welch changed General Electric from a sleeping giant to a top competitive organization and then kept them at the top by adopting a "boundaryless organization" in which the employees adopted "boundaryless behavior" or did not stay employed, is recorded in several book on the subject.

Ken suggested that there is a mechanical way to think about an organization, and then there's a biological view. He gave GE Capital as an example. A traditional structure is a hierarchical approach, which suits a line-of-responsibility and authority command chain within an organization but does not accommodate the new style of management which embraces relationships with customers, with employees, with business partners, and with time and space (integrating work into whole life).

Success factors used to be measured as size, role clarification, specialization, and control. The new success factors are more about speed, flexibility, integration, and innovation. In this new environment for business, preoccupation with structure is not the answer.

Ken identified four types of boundaries in an organization:

• Vertical
• Horizontal
• External
• Geographic

and four organizational levers:

• Information
• Capability
• Authority
• Rewards

explaining that the challenge is to integrate these levers. The most powerful way to achieve this integration is through dialog. Once started, his experience is that this sense of a boundaryless organization escalates to become a natural part of everyday work.

Top

Continue to Tuesday sessions

Home · Contacts · Legal · Copyright · Members · News

© The Open Group 1995-2012  Updated on Friday, 22 August 2003