The main challenges of doing Grid-based business

This article is based carried out by the BEinGRID project which ran 25 grid business pilots between June 2006 and November 2009.

The increasing need of powerful computing resources, along with the ability to manage, distribute, process and store large amounts of data, has motivated the development of Grid technology over the last 10 years, particularly in the scientific area, but also in some commercial and industrial areas, e.g. the finance and pharmaceutical sectors.

While this process has made Grid computing into the state-of-the-art technology of modern computing, the world is still waiting for Grid technology to enter the main stream IT services market. The statement made by Ian Foster in 2004 is still true: “I think it's inevitable that Grid computing - and the virtualization and federation of resources - will happen. Less clear to me is how quickly people will accept the standards that are being set and begin to demand Grid computing”. Of course, people will not demand Grid computing per se. Instead they will demand products and services that can best be delivered by Grid computing, e.g. SaaS (Software as a Service).

This article summarizes the most significant challenges of Grid based business and tries to answer the question: which main challenges can keep a Grid based business from being successful?

Challenge 1: Market need
Every new business has to check and verify very carefully, if the product and/or service meets the expectations and needs of the targeted market. In a nutshell: if there isn’t an unsatisfied market need, there isn’t a business opportunity. Most BEinGRID pilots looked at in this research took consideration of this crucial key success factor by integrating one of their first users into the project to provide customer focus.

An example from BEinGRID pilot number 8: “Another risk (in general) is the development of software that is far away from end user’s and market requirements. Due to close cooperation with the pilot user FIUKA – even during development stage – this risk can be minimized. So, it is guaranteed, that the developed software solution meets the market requirements.”
While this strategy is definitely helpful to adopt the solution to that specific customer’s needs, future service/solution providers have to be careful to extrapolate: meeting one customer’s needs is not a guarantee that the product and/or service will meet the needs of the whole market.

In general it can be said that most of the markets the BEinGRID pilots have focused on and which do not (yet) use Grid technology in their business, may not have a immediate need for Grid technology as such. However, they may develop a need for applications that can best be delivered over a Grid platform, especially if these applications will add new features to their existing product portfolio or even offer completely new business opportunities.

This leads to a common misunderstanding when introducing a new technology into a
customer’s organization. The misunderstanding is that the introduction of new technologies is often addressed to the customer’s IT organization instead of addressing the introduction of new services (which require the new technology) to the customer’s business organization. In other words: as long as the Grid based services do not add significant value to the customer’s core business, the customer’s IT organization won’t get the budgetary authority to introduce the new technology.

Every potential service provider should be aware of the four general business and IT roles within a customer’s organization:
•    The business architect, who defines the services on which the customer’s business relies on; he also defines the requirements for the IT services.
•    The business people, who actually run the customer’s business.
•    The IT architect, who defines the IT architecture and the IT services on basis of the requirements of the business architect.
•    The IT people, who actually develop, integrate and run the IT infrastructure.

Considering those roles it is obvious that the introduction of new services can only be initiated by the business architect, hence the offered Grid based service has to offer new benefits to the customer’s core business. In a nutshell: no customer will introduce Grid technology just because it’s such a great new technology.

It has to be added that many customer’s organizations are on their way to move into service oriented organizations with the roles described above, but they’re not there yet. This applies to the business aspects as well as to the IT aspects. This makes it even more difficult for potential service providers to address the decision makers in a customer’s organization.

Challenge 2: Reluctance to adopt Grid based services
Several BEinGRID pilots stated in their Exploitation Plans that the specific market is reluctant as regards to Grid technology, because the market is not aware of and not familiar with Grid technology or does not feel confident with certain issues of Grid technology, such as reliability, security, stability, maturity, etc.

There are several reasons for this reluctance. First, the adoption of Grid based services goes hand in hand with modern software and hardware architectures, which means that there is an obvious correlation between the knowledge and adoption of Grid based technologies and familiarity with and use of SOA (Service Oriented Architecture). It does not matter if the customer plans to use Grid based services provided by a service provider or if the customer plans to build up a Grid based infrastructure himself – in any case the customer will have to adapt his IT strategy to service based technology in order to use Grid based services and (even more important) to integrate Grid based services with the rest of the IT infrastructure and legacy applications. From a customer’s point of view this results in a transition process with significant technological and cultural changes. Keeping in mind that the customer’s IT personnel works under operational pressure with limited budgets, skills and capacities, it is quite understandable, that the majority of the companies do not jump on the Grid technology bandwagon without a lot of thought and careful preparation.

Second, the overall data management gets more complex. While it’s already a complex task to maintain consistent and safe data management within the organization, this gets even more complex if part of the crucial company’s data is processed and stored outside of the organization. The upcoming data security and data integrity issues have to be very clearly solved and defined. At the end of the day, these issues affect the overall information lifecycle management of an organization.

Third, the use of Grid based services goes along with a new quality of contractual agreements (eg SLAs, SaaS, etc) between the customer and the service provider. Depending on the provided service the accounting and billing of services gets more complex. It has also to be kept in mind that external costs are usually billed to internal parts of the organization depending on the use of services; a different external contractual agreement will affect the internal billing structures as well.

The Business Experiments identified several aspects of this challenge in their Major Risk Analysis / SWOT Analysis, such as:
•    BEinGRID pilot 03 “The Business Experiment relies on SaaS which is a new way to use software. It takes time for customers to switch to a new business model.”,
•    BEinGRID pilot.04 “Security issues and reliability are two factors which are fundamental in the financial sector. A low performance on these two factors may lead to a rejection of grid applications from the actors in the financial market.”,
•    BEinGRID pilot 06 “We are dealing with a completely new technology in the groundwater modelling environment. It will be difficult to convince the clients from the new service.”, BEinGRID pilot08: “Main concerns on reliability from an end user’s perspective are a lack of stability and security…”
•    BEinGRID pilot 18: “Some of the potential clients may prefer to build their own computing resources centres for their IT needs instead of adopting our EGEODE (Expanding Geosciences On Demand) service. There are some good reasons for this: fear of IPR, data security, data exchanges, and security issues. They could be very conservative or nationalistic about questions concerning data processing or data storage sites. They could also be reluctant to change their habits. The sales and marketing team will have to work hard to change such attitudes.”

A service provider for Grid based services has to be aware of the reasons for the reluctance of his potential customers and in order to be successful these issues have to be addressed in the market approach. Only those service providers who support their customers in going through this transition process to use Grid based services and make it as easy for the customer to hop on the Grid train will be successful on long term.

Challenge 3: The innovation cycle
A crucial aspect to consider for potential providers of Grid based services is the innovation cycle. Grid technology is innovative for most markets. The same applies to Grid based services. According to Geoffrey A. Moore there are four phases of evolution of IT capabilities:
•    Invention: enabling activities have to make sure that customer’s organisations are able to use a new IT capability.
•    Innovation: new innovative services based on new IT capabilities will be established; the main focus is differentiation among competitors.
•    Standardization: the majority of players in a specific market segment will adopt the new IT capability; a bandwagon effect takes place, the IT capability becomes a de facto standard; the main focus is productivity now.
•    Commoditization: the IT capability will be a simple commodity from the point of view of the customer, hence are the services based on this capability; the market gets high  competitive.
On top of that, the different phases of the evolution of IT capabilities are represented in the market by different types of customers: technology enthusiasts, visionaries, pragmatists (early majority), conservatives (late majority) and sceptics. According to Moore, the most challenging step is making the transition between visionaries and pragmatists (early majority), hence doing innovative business (such as Grid based business) in a mainstream market.

It is obvious that the introduction of Grid based services is strongly connected with the
introduction of SOA, a process which finds itself in a more or less lingering state over the last 5 years. Many customers are still on their way to move into a service based organization, which makes use of service oriented IT architectures, hence they’re still on their way to “be enabled” for Grid based services. This means that it will take time for the new technologies (here: Grid based services) to achieve commercial grade performance and productivity levels - definitely from the customer’s and market’s point of view, and also for many potential providers of Grid based services as well.
Every potential service provider has to qualify a.) in which phases of the innovation cycle he can do business b.) how he can survive the long wait until his business achieves lift off c.) how he can support his potential customers to move quickly through the first phases Several Business Experiments have a clearly defined market entry strategy for their grid based solutions, for example BEinGRID pilot 03 (3D rendering and real-time visualization), BEinGRID pilot 04 (financial portfolio optimization), BEinGRID pilot18 (seismic simulation) and others. Those strategies mainly refer to the market’s early learning phase. The main goal of this phase is to make the market familiar with grid based solutions, to achieve short term ROI and to win as many potential customers as possible. The impact of the complete innovation cycle on grid based business has not been in the focus of the Business Experiments though. Still it’s a key success factor for sustainable business.

Challenge 4: Unclear ROI
Given a product/solution which satisfies the customer’s business needs, and given a customer who is willing to go through a transition process to make use of Grid based services, most companies are still unconvinced about the Grid’s ROI potential. According to a 451 Group report published in 2004 “much work is still required to convince the most risk adverse majority of users across all verticals that Grid investments will pay dividends”. Although this report was published in 2004 this question lingers on - and there is reason to be cautious. There is still uncertainty about the numbers. Companies need to differentiate between the hype, of which there is much, and hard facts, of which there is much less. Things like this need to be worked through rigorously in detail. If it’s not done thoroughly, the chances of failure are much increased. Wasting money is easy, saving money is hard, making money is harder still. And, getting it wrong could be a career limiting move.

Challenge 5: Unreliable technical and business premises
According to “Fallacies of Distributed Computing Explained” by Arnon Rotem-Gal-Oz there are eight fallacies of distributed computing:

      1. The network is reliable
      2. The latency is zero
      3. The bandwidth is infinite
      4. The network is secure
      5. The topology doesn't change
      6. There is (only) one administrator
      7. The transport cost is zero
      8. The network is homogeneous

No competent businessman will want to risk falling foul of one (or more) of these fallacies so they will want to make sure than none of them can and/or will undermine their Grid based project. For example, the seventh fallacy of distributed computing is: “the transport cost is zero”. According to a paper published by Jim Gray of Microsoft Research in 2003, the transport cost can make some distributed computing projects uneconomic. If the principles explained in Mr Gray’s paper are still valid, then the cost savings generated by Grid computing can be more than offset by the cost of moving the data to and from the computing resource. If so, then co-locating the hardware with the data ought to be more cost effective than sending the data to and from remote hardware for processing. And, because computer processing power is decreasing in cost more rapidly than transmission bandwidth, the cost benefit analysis is unlikely to be moving in favour of remote computing.
If so, this may mean that Grid computing will only be best for applications where the advantages out weigh the disadvantages. For Grid computing to succeed it has to tick as many of these boxes as it can. And the ticks must be based on verifiable facts not wishful thinking. Clearly, the result will depend on the particular circumstances of each project.

Each project will have to prove it offers the best cost benefit outcome against alternatives that are also trying to prove that they offer the best cost benefit outcome. If Mr Gray is right, the cost savings offered by Grid computing can be more than offset by the cost of moving the data to and from the computing resource. If the cost of moving the data to and from the computing resource is a key cost element, then Grid applications that can circumvent this obstacle can still gain traction.

The question is: what might these circumstances be?
1. Free transport
If somebody claims that data can be moved about at no cost then you should consider the possibility that, by accident or design, they are not being accurate. As a general rule, there is no such thing as a free lunch - somebody somewhere has to pick up the tab. What they probably mean is low cost transport.

2. Low cost transport
For example, if a company owns its own network, and the data will not travel outside this network, then the cost of sending the data over the network may be considered to be “free”, but it’s unlikely to be completely “free”. Even if the company owns the network, the network will have capital and operating costs. Putting extra traffic over that network (if the network has sufficient spare capacity) will incur a marginal cost (ie increase operating costs by something) and if the network does not have sufficient spare capacity, new capital expenditure could be required.
Most internal networks are cost centres and they often recover their operating costs by internal charges. These should be less than an externally sourced cost, and they may just be a cost allocation, but they still represent a quantifiable cost. Of course, the most obvious companies that own their own networks are telcos, but their business model is based on charging users for using their networks, so they have no incentive to offer transport (even to internal customers) at less than the market rate.

Notwithstanding the above, already having a network with spare capacity where marginal costing can be applied, and/or having access to cheap connectivity, ought to give companies an edge if they wish to use Grid based applications and services themselves and/or offer Grid based applications and services to external customers. Even if a company has its own network (say) made up of leased lines, some kinds of Grid computing mean sending the data outside the enterprise for processing so this advantage may not prove to be that much of an advantage. Alternatively the additional data transfer may fit within the allocation the company already pays for (say) under a flat charging scheme. If so, the extra transport cost could be little or nothing but if a company has plenty of spare capacity on its external communications links one could argue that it has been paying for more capacity than it really needs and it may have a cost reduction opportunity if it does not use the spare capacity.

3. Limited data transfer
If the amount of data to be sent over the network is limited but the processing required is large, then the cost of sending a little data may be less than the savings generated by doing a large amount of processing remotely. This will also apply if a company plans to provide a Grid based SaaS over the Internet. The service would have to be something that potential users could use without running up large bills for telecommunications services.

4. Intermittent demand
If a company only has a part time need for lots of computing power, then the cost of sending the data over the network may be acceptable, if the total cost is less an alternative solution. This, of course, is the classic telco business model where the telco builds an expensive infrastructure and then lets many customers make multiple use of it for multiple small charges. The sum of many small charges adds up to enough to provide a ROI on the original investment. Because this is the way telcos already do business, they are ideally placed to provide distributed computing services because distributed computing services are a logical add-on to their existing service orientated business and compatible with their existing business model, i.e. their billing systems are designed to cope with usage based charging.

For this reason, this is perhaps the most likely scenario. If a company has a need for demanding computing 24/7 then it will probably be more cost effective for the company to build and maintain its own data centre. But if the need is only occasional, even if it is regular, then opting for a bought in service will probably be more cost effective because the company would only have to pay online charges when it was using the service. The cost per hour may be more - but the total cost over a period of time should be less.

Case Study – BEinGRID pilot 18> The seismic imaging and reservoir simulation service requires the processing of large amounts of data collected in often remote locations around the world. The pilot recognised that sending such large amounts of data to them over a commercial network could be very expensive for their potential customers. Therefore, they also geared up to receive data for processing on storage media sent via international courier services before doing the processing itself on a Grid. In this case, the pilot has thought through how their potential customers will use the service, what problems they might face, and put in place a solution. This is a very good example of a BEinGRID pilot doing “a walk through” to check for impediments to the business plan. This sort of attention to detail is what often separates successful companies from those that fail. Like any business, the BEinGRID pilots need to
make sure that buying and using their product or service is as simple as possible for the potential users.
This section has focussed on “the transport cost is zero” but this is only one of the eight fallacies of distributed computing. To be successful, Grid based businesses need to verify that none of these fallacies will be an impediment to their business plan. In 1995, Deloitte and Touche surveyed a large number of CIOs (chief information officer) to find out why new and innovative projects had failed. Most commonly quoted barriers to change were change resistance, lack of sponsorship, unrealistic expectations, poor management, uncompelling business case, lack of skills, project scope uncertainty, lack of change management plan, lack of holistic process view, and non-integration of IT perspective. Sadly, it seems that grid computing probably faces exactly the same problems.