Unified ecosystems
To enable a wider range of more demanding use cases, the network compute fabric needs to facilitate the transformation of ecosystem engagements. This will be realized through the collaboration of a broad set of actors, driven by the need to mutualize the cost of infrastructure, offer services at larger scale and enable full roaming between providers. Network and cloud providers, application developers, service providers, and device and equipment vendors all have a role to play. Network operators have an opportunity to utilize their distributed network infrastructure as an innovation platform for richer service offerings, combining connectivity with compute in both wide-area and on-premises edge deployment scenarios [4].
Several ecosystem models may evolve and have already been described in the context of edge computing [5]. There are four basic models for network compute fabric services: standalone, facilitation, aggregation and federation. The standalone model is where the network compute fabric services are offered by one actor in a standalone manner. The facilitation model is where one provider acts to unify fragmented offerings from other providers. The aggregation model is where one provider is active from a business perspective in building and reselling bundles of services from multiple providers. The federation model is when the providers collaborate within a framework to utilize each other’s offerings.
Ecosystem partnerships involve dynamic business agreements that need to be reached regarding the usage of resources between multiple partners. They also involve technical challenges of integrating services from the different actors. Ecosystem partnering can be facilitated by standards that ensure interoperability or by technologies that automate the handling of partner relationships. The right level of harmonization in the ecosystem is key to support scalability as well as innovation.
Unified application management
The network compute fabric will be a highly distributed platform that enables the execution of applications across multiple administrative domains. This innovation platform will require intelligent and data-driven operations to seamlessly span devices and network and cloud domains. An integrated DevOps toolchain will enable the appliance of similar development and operational methodologies across those domains. By separating orchestration from application functionality, the network compute fabric will allow tailored optimization for different network domains without the need for application code change.
Applications are usually composed of one or more services that may have different deployment and performance requirements. With interoperable abstractions and programming models, services can be easily deployed across different domains. With common operation models and tools, service quality can be assured autonomously per domain, considering joint network and compute resources and domain-specific performance requirements. Applications can also be collocated with other applications with dependent performance requirements. Common optimization loops will then enable operational synergies across such applications – that is, data-driven shared knowledge can proactively optimize dependent applications beyond individual optimization models.
Generalized operational pipelines and tools with programmable infrastructure will also enable the offloading of the common orchestration functionality from applications to the fabric. This will enable simpler application programming models with a stronger focus on the core application functionality.
With common data pipelines and data governance models including data protection, data flows can be combined across resource, functional and administrative domains for multi-layer data classifications that can bring aggregated and more accurate optimizations beyond single domain maturity.
Unified execution environment
The unified execution environment will act as an operating system, providing fundamental functionalities and services on top of distributed and heterogeneous network, compute and storage assets [6]. Evolving the ideas of serverless computing, it will facilitate the development and deployment of distributed applications on top of this infrastructure. The application has access to a compute service that always appears local, despite dynamic network changes or user/data mobility events. The unified execution environment will simplify the development of distributed applications by offering several capabilities.
Seamless task mobility will be enabled by evolved container formats based on portable bytecode such as WebAssembly and associated system interfaces. This will allow the platform to dynamically schedule workloads on nodes regardless of varying hardware and system software setups. As a result, several optimizations can be performed with limited overhead, such as moving computations close to a data source or consumer. This can be useful for a variety of reasons, including having computational tasks follow mobile users, offloading workloads from the device to preserve device energy and moving computations for an optimized cost/performance trade-off.
A distributed data infrastructure with a range of built-in consistency mechanisms will allow developers to choose from the right-sized trade-off between consistency levels and their associated resource costs. This will offer application tasks seamless access to data stored in multiple locations, as if from a single local access. As the application components will be distributed between several locations in a pipeline fashion, the platform would be in full control of both the compute and data locations, as well as traffic management and prioritization. As a result, data access can be optimized along the computation pipeline even considering network characteristics and usage profiles, while obliging legal or operational data policies.
Unified exposure of network and compute capabilities
The computational environment in the network compute fabric will be heterogeneous, which will increase with emerging computational hardware complementing traditional network equipment. One example is hardware acceleration technologies that provide compute and storage capabilities specialized for certain workload tasks, such as graphics processing units, field-programmable gate arrays and storage-class persistent memories.
By adapting tools and languages like OneAPI [7], developers will be able to write a single-source implementation of an algorithm that will be able to run on different hardware choices. Selection of actual processing or memory/storage instance will be based on availability and performance requirements, making it easy for developers to create performant applications that can run anywhere.
The network compute fabric also opens up for extensive in-network computation. Modern transport networking equipment will no longer be limited to packet transport, but also able to provide programmable compute capabilities, opening up for new divisions of application functionality where some tasks, such as QoS, scheduling, policing, encoding and recoding can be performed directly in the data plane.
The network compute fabric will enable efficiency gains through transparent shortcuts inside the infrastructure between application components and the network and compute platform. Once applications are collocated with network functions on the same host, rack or cluster, parts of the network and operating system stacks can be bypassed. Depending on the situation, different technologies can be used, such as modern interconnect technologies and protocols in combination with fast replication services like Derecho [8].
"network" - Google News
June 30, 2021 at 03:12PM
https://ift.tt/3duDyvp
The network compute fabric – an integral part of 6G - Ericsson
"network" - Google News
https://ift.tt/2v9ojEM
https://ift.tt/2KVQLik
Bagikan Berita Ini
0 Response to "The network compute fabric – an integral part of 6G - Ericsson"
Post a Comment