Application of a multi-sourcing strategy
By Axel Winter
25 May 2009
During my time in the industry, one of my tasks was to set up a global core banking platform, as a shared service centrally. For this to happen, I also had to manage the setup of two global data centres, with the plan to evolve these towards a global application hosting centre, for service-oriented architecture-based application frameworks. At the time I pursued a multi-sourcing strategy, with project management, enterprise architecture as well as technology subject matter expertise, provided internally, to complement the service provider skills, although the vendors have been multinationals in their own right.
In modern terms, this was an internal ‘banking cloud’. Next to the standard implementation issues, in such engagements one of the key challenges was that this specific version of software had limitations on hosting ‘multi-entities’ within one instance of the software. Now the trick, of course, in cloud computing, is to have as few instances as possible, to manage cost and speed of implementation. Hence we teamed up with the vendor, to address some of these shortcomings and create a better multi-entity platform.
The initial phases meant to get the data centre, servers, and networking ready. I still recall my frustration in working with professional outsourcers and the time and discussion it took to get just an absolute standard infrastructure platform ready: A virtualised mid-range environment with different operating systems, database and Java application server software, plus things such as monitoring, backup, site replication, and basic technical operating processes.
‘Ready in minutes’ goal
What struck me most is that I would have expected the mid-range portion to be ready in minutes, and that it should have taken a clean install of all software elements and the banking solution in a few hours. Until today, I am hearing from global infrastructure providers that duplicating a virtual machine, inclusive of all process steps, can take up to one week. It’s important to understand that basic computer and storage resources, with standard software, are really a commodity setup and can easily be replicated. Hence my goal of ‘ready in minutes’ is actually an achievable one.
The reason for such delays is that we are managing commodity items still like a ‘customised’ product and not like a utility, even if marketing wants us to believe otherwise. Our overall model in IT is focused on a customised way of doing things and not of replicating the best practice.
Few early entrants in the service provider space are proving that ‘ready in minutes’ is actually possible and by using technologies available today. These companies are introducing slowly, but steadily the next generation of IT to us: IT3.0 – The Industrialised Information Technology (iIT), which is about knowledge workers and less about ‘manual labour’. Readers well-versed with the industrialisation age 100 years ago should actually find the parallels rather interesting.
Global virtual data centre
Today Amazon.Com enables you to use a Web service and a credit card to buy storage and computer facilities. In effect, you could automate the end-to-end process of increasing and adding such resources, to your ‘global virtual data centre’.
Now, consider what Amazon.com must have done to generate this global department store. The internal environment must be incredibly mature and they managed to industrialise their infrastructure, provisioning and other processes, to resell this as a service reducing their total cost of ownership (TCO).
Effectively, every company can do this and, of course, it would be highly expected from service providers. Creating such a technically competent people, and process maturity takes, on one hand, a bottom-up approach, to evolve capabilities via several iterations and, on the other hand, clear goal setting, of a smart CTO, like Werner Vogels (CTO at Amazon.com).
Application delivery advances
We are not only seeing great advances within the infrastructure space towards virtualisation, process control and service enablement, but also in the application delivery space:
Web 3.0 is being called the Semantic Web and is probably best described by Tim Berners Lee's (the original HTTP aka Web 1.0 inventor) in his invention ‘Linked Data’: Essentially having an easy data definition engine, enabling the linking related data elements together (check http://www4.wiwiss.fu-berlin.de/bizer/pub/lod-datasets_2009-03-05.html for a visual linked data browser).
This is actually a crucial development. It’s not important, whether Linked Data will succeed as a protocol standard, but it shows the amount of industrialisation within the data processing and software layer. In the same way, we can write mechanisms to automatically add storage and computing power, we can write methods to query data regardless of their source and use this data according to the value they represent to us.
The next level of industrialisation happens within the software development itself. We are seeing with new tools (open source or commercial), that we are able to draw processes and use drag and drop windows editing, to paint user interfaces tying back to prebuilt (service) logic. The industrialised way and logical next step is to take these models and attempt to create source code or Web service definitions.
Model-driven architecture
The technology and approach for this is called ‘Model Driven Architecture’. This essentially means there is a ‘drawn’ model (similar to process charts), which contains also configuration items. These models can be aligned to pre-built source code or service artifacts, which can be linked by workflows.
Some of these tools can already be used by ‘power users’. Right now, these tools are still in their early stages and either require a lot of manual work to create process flows or ‘models’ or the technology is highly proprietary in nature, preventing true uptakes and yet other approaches are not accurate. As a result, these frameworks are still only being used in targeted deliveries.
The goal on software development still is to evolve model-driven architectures and drive higher productivity (time to market + quality) in the delivery cycles, something only being realised by stringed delivery methods generating overheads.
Even so, there are still arguments in the media whether we will see virtualisation technologies, services architectures (SOA), Web 3.0, etc. These things are being used today and only ‘Power Point experts’ call this immature. Open the hoods of your ERP or banking environment and in many cases you will see full service architectures or at least a layer or wrapper around their legacy, enabling the adoption of service-based interfacing.
Merging application silos
At the Asian Banker Summit on 10 and 11 May this year, I predicted that based on these changes, we will see finally a merger of application silos within the IT organisation and create a single application team with different functional expertise. Meaning, if everything is a service, we would not need different teams around application boundaries anymore, but more functional and technical expertise.
The logical conclusion of this forced reorganisation in modern IT departments will be reorganisation within the business landscape as well. How can you distinguish call centre operators by products or functions, if a sales guy nowadays will need the same capabilities as a collector and both are most likely getting their ‘task orders’ from the screen?
During my visit at the Open Group Conference in London, we had some interesting discussions on how enterprise architecture will evolve until 2019. Naturally, there have been different views around this, but essentially we all agreed on the following themes:
• Reinvention of corporate planning and strategy – merging of IT and business planning approaches
• Driving solution design for the so called ‘network enterprise’ – an enterprise which uses a multitude of partners, business and IT services to offer capabilities to clients
• Creation of a ‘chief strategy officer’ owning the planning component across all competencies in an enterprise
• Business simulation and scenario-based planning to become a new core competence, with IT being but one part in this.
Increasingly complex
Now, this does not mean that IT will become simpler; the opposite is true. IT will become more complex. Specialised technology and business skills will be more important than ever, but it will generate more value and more power for the end-user or customer, with more productivity in less time.
The key characteristics of IT3.0 will be the following:
• Business simulations will drive detailed IT planning, which will use prebuilt components to construct capabilities for the business
• These prebuilt components will be delivered by a set of internal and external service providers.
• With IT moving to centrestage, it will become a much more ‘unique differentiator’ for business and hence true business innovation will happen internally by IT
• This trend will drive specialisation of IT staff and kick off an avalanche, increasing the business and technology skills across all layers of the IT delivery and operation value chain
• Multi-year projects will be an experience of the past and typical projects will take three months, with really long projects being 9-12 months
• This is driven by essentially reusing prebuilt or industrialised components to 50, 70, or 80 per cent for new capabilities.
• This is true for infrastructure as well. We will see global hardware and outsourcing providers quickly adopting the Amazon.com model and perfecting it and making it work for more complex requirements.
• For outsourcing, the ‘jury is still out’. It will largely depend on a) value generation (versus price alone) and b) the ability of corporation to retain a unique IT differentiator within this model.
IT3.0 core decision-making factor
What provides the corporation with a unique sales proposition versus a ‘me too’ approach? The ‘me too’ part will be outsourced and provided by public clouds.
Is there something I should do today, to move into the right direction?
Right now for most organisations, the focus should be to increase the skills of IT staff for application delivery, infrastructure, and associated methodologies. Additional business staff also need to develop a better understanding of the complexities and opportunities in dealing with IT environments.
The goal for the enterprise architecture is to become more flexible and dynamic and hence increase the speed of delivering X fold. I would argue this is a universal need for most businesses in the future, even for business departments.
Flexibility is driven by all layers of IT: people, processes and technologies. The challenge is to see the dependencies between application and infrastructure layer. The infrastructure should focus on a standard operation systems and hardware model, using a strong virtualisation approach. Virtualisation here will allow a much easier and hardware independent creation of ‘packages’ around databases, application servers, and even business applications such as ERP packages, generating a vendor situation, which caters for instant delivery of additional storage and hardware resources. This will create the ability to almost instantly generate additional productions, disaster recovery or test environments.
Separate evolution
The application tier, will most likely drive towards a service-oriented architecture and the utilisation of an enterprise service bus will allow for generating layers of abstraction between distinct areas of enterprise architecture. These can then be evolved separately from each other, assuming proper data and interface design work.
Critics may argue that this is neither easy nor cheap. I beg to differ. Actually, this will seriously lower the IT spending, by providing additional capabilities. Consider this:
1.) Setting this goal will allow IT teams to upgrade their skills, which will enable a different approach between service provider and client.
2.) Open source platforms and different infrastructure models (virtualisation, low-end platform hosting models, information lifecycle storage cost reduction mechanism, etc.), will drive costs down and increase the adherence to industry standards
3.) Introduce ‘lean IT methodologies’ to drive down IT process and cycle costs significantly, by increasing quality, driven by strong content-based quality assurance processes.
This article is meant to give the reader a good view on the current and future framework. The next step is to look deeper into current and future approaches to application and infrastructure architecture and their alignment with a new concept called business architecture. This will be the topic of next article on ‘lean enterprise architecture’.
Axel Winter is the Director for IT Consulting at Deloitte Consulting SE Asia.
