The cloud has been going through a continuous evolution process since its birth, and each year we witness highly relevant advances. The fast rise and adoption of container technologies was one of the most visible developments in 2014.
Although 2015 is still in its early stage, we can already identify some trends that are starting to see the light of day. Among these is what we can call the “disaggregation” of the cloud – one of the most fundamental evolutionary steps of the cloud in recent years. This evolution will unlock a set of all-new possibilities, from an operational as well as from a service perspective.
But what do we actually mean by “cloud disaggregation”? We discuss disaggregation on two levels: the hardware level and the facility level.
The former refers to hardware resource disaggregation, where the long lasting and traditional server architecture boundaries are broken up and hardware resources become independent units. The most tangible realization of this disaggregation today can be seen in Intel’s Rack Scale Architecture vision.
The latter one, disaggregation at the facility level, relates to the well-known distributed cloud concept. While the distributed cloud is considered an undeniable step since some time now, there might be a few doubts about how distributed it will actually be. However, we’re starting to see some concrete cloud distribution patterns.
We in Ericsson have been discussing our distributed cloud vision for several years. At the Mobile World Congress 2015, we released the first Hyperscale Data Center System built on Intel Rack Scale Architecture – an important step on the hardware disaggregation journey.
For the third year, we at Ericsson Research brought demos to illustrate visions and solutions to come:
- Capillary Networks: how to manage large-scale IoT networks and make it possible to handle low latency control loops using the mobile network.
- Knowledge management for industry and society: using intelligent transport as the example, it showed how freight train transportation use cases can be handled by automation and knowledge management.
- Network Near Analytics: how distributed analytics procedures at the edge of networks can improve its overall management.
- Cloud Evolution towards 5G: showing evolved cloud architecture to use cloud flexibility and automation to support NFV deployment and management, reduce OPEX and TTM.
How are these above-mentioned demos related to cloud disaggregation? For one thing, they all require that cloud capacity be distributed in the network – in some cases with higher capacity, in others with lower. Moreover, there is the facility size diversity, where sometimes you will need a small or mid-size facility, and sometimes a simple box will do the job.
At first sight, the technological impact of cloud disaggregation may seem confined to the hardware layer and to the increase of cloud facilities. However, the impact goes further. Currently deployed cloud management software, sometimes referred to as cloud operating systems (for example OpenStack), is neither prepared to support a highly distributed cloud, nor to manage such fine grained sets of individual hardware resources. And by “support” we mean everything: scalability, reliability, integrity, security, performance, QoS, governance, and many other aspects.
Cloud disaggregation is happening at both hardware and facility levels, its impact has already been noticed, and the development will continue. Our team at Ericsson Research will stay focused on clearing the road to the disaggregated cloud environment. As part of our work we will keep contributing to the build-up of a thriving research community in the area – for example by organizing workshops.
João Soares, Ericsson Research