The decentralised approach For the past 20 years , data teams have typically focused on collecting data from across the enterprise and storing this information centrally in either a data warehouse , data lake , or repository , according to Ryan Moore , Head of Data and Analytics at Aiimi , a creative tech company that specialises in artificial intelligence ( AI ) and data :
“ While this centralised approach has its benefits , such as increased accessibility , as well as traditional data modelling and reporting capabilities , it also moves data away from its natural source and from the teams who need to quickly derive insights relevant to their bespoke requirements .”
Moore suggests that rather than rearchitecting infrastructure to align with a centralised data strategy , an alternative is to adopt a decentralised approach — such as a data mesh — that interconnects all information , including structured and unstructured data , across the enterprise and enables the self-servicing of data and analytics capabilities for each business domain .
“ By moving to a data mesh model , in which data is organised and owned by each business domain , bespoke data products that conform to standardised data principles can be delivered with ease . Teams can also scale their data requirements independently , delivering increased business agility and value ,” says Moore .
To centralise or decentralise ? Martin Molloy , Partner , CIO Advisory at consultancy KPMG UK , says in recent times , decentralisation has also started to happen naturally . As technology has become more accessible through Cloud , there has been greater adoption of technology outside of IT than ever before . This used to be referred to as “ shadow ” IT but in reality , Molloy suggests it is just businesses using technology to solve problems .
“ The reality is , this is never a binary decision . Most clients we work with rarely have a completely centralised or decentralised model , there is almost always a hybrid somewhere along the spectrum ,” he said .
technologymagazine . com 97