Implementing the NetroSphere Concept at NTT

1.1 Configuration

NTT Network Service Systems Laboratories is developing a new server architecture called MAGONIA^*1 to enable the early creation of network services and greatly reduce system development and operating costs as part of the NetroSphere concept. The conventional approach to developing network systems is to use hardware and software optimized for each network function while meeting stringent requirements in reliability and performance that are demanded of a carrier. However, a system having such a separately optimized silo-type equipment configuration increases the quantity and types of necessary equipment while requiring operators to be located separately. This type of configuration consequently drives up operating costs and keeps maintenance and management costs over the entire system high, and as such, it has hindered the development and rollout of new services.

MAGONIA configures network functions as software separate from hardware and provides a common platform for efficient development of those functions as applications (Fig. 1). The development of network functions using MAGONIA makes it easy to ensure scalability and high reliability, which in the past could not be done without driving up development costs. In particular, using MAGONIA to achieve scalability enables small starts for services and facilitates the early creation of new services.

Fig. 1. MAGONIA architecture.

1.2 Platform functions

MAGONIA features two platform functions: a distributed processing base for achieving an n-active (N-ACT) server cluster and a virtualization base for achieving a resource pool (Fig. 2).

Fig. 2. MAGONIA’s distributed processing base and virtualization base.

The distributed processing base enables the construction of an efficient N-ACT cluster in which all servers are up and running by having the active system (in which processing is in progress) replicate the data of other servers, thereby doubling as a standby system. This negates the need for dedicated standby servers as in the active and standby (ACT-SBY) architecture used in conventional network systems. In an N-ACT cluster, servers are made to store replicated data based on a sorting algorithm called consistent hashing, which enables the number of servers making up a cluster to be increased or decreased on the fly without service downtime. In this way, system operation can always be achieved with the least number of servers needed to meet demand.

A resource pool, meanwhile, acts as a common source of resources (e.g., the processing power of central processing units (CPUs), the recording capacity of memory, and storage on physical servers) for multiple systems providing them. The resource pool in MAGONIA is achieved by the virtualization base, and it allocates resources in the form of virtual machines (VMs) according to requests issued by applications making up each system. This virtualization base conforms to the architecture framework of the European Telecommunications Standards Institute’s Industry Specification Group on Network Functions Virtualization (ETSI NFV ISG)^*2 and consists mainly of the network function virtualization infrastructure (NFVI) based on VMs and a virtual network function manager (VNFM) that provides a management mechanism for autonomous control of those VMs (Fig. 3).

Fig. 3. Correspondence between NFV architecture and virtualization base.

A system applying MAGONIA using these platform functions can increase its facility usage efficiency by a significant amount compared to existing systems. Furthermore, in the event of a server failure, it can incorporate an alternate VM into its N-ACT cluster from the resource pool, thereby enabling autonomous recovery without having to dispatch maintenance personnel to the site where the failure occurred. A reduction in system operating costs can therefore be expected.

At NTT, we are promoting more open activities so that a variety of companies and enterprises can participate in the implementation of MAGONIA.

1.3 Release of API specifications

In this regard, we envision the distributed processing base to be provided as middleware. In February 2015, we released the specifications for the application programming interface (API) of the distributed processing base so that anyone will be able to design middleware and applications conforming to those specifications.

As for the virtualization base, there has been rapid growth in cloud-related commercially available virtualization techniques in recent years, and at the same time, NFV-related products have been regularly launched on the market with the progress in standardization of NFV as virtualization architecture for network systems. We can expect competition in NFV-compliant products among a variety of vendors to intensify in the years to come. Our plan is to expand the use of the distributed processing base in combination with the virtualization base with the aim of implementing MAGONIA and achieving early implementation of the NetroSphere concept.

2.1 Optimized use of resources

The NetroSphere concept envisions a low-cost network made possible by combining common components. This will require technology for making effective use of pre-existing resources, which differs from the conventional approach to facility design and operation in which the service to be provided is known in advance and the facilities required for the service are deployed appropriately. Resource optimization technology aims to meet this need for using pre-existing resources by determining an optimal combination of software and hardware functions unevenly distributed in the network and implementing them on the physical network via integrated control functions (Fig. 4).

Fig. 4. Network resource optimization based on proactive control.

2.2 Key technologies

Specifically, the introduction of virtualization technology typified by software-defined networking (SDN) and NFV will enable the definition and construction of services on the logical network independent of the actual configuration of the physical network. The use of virtualization technology and the technologies summarized below will enable individual services to be mapped to the physical network in a flexible manner.

(1) Traffic prediction technology considering traffic-generation mechanisms

In addition to conventional techniques that uncover trends based on traffic observations, this technology will use human-flow analysis in physical space to reflect people—the source of traffic—and terminal mobility in traffic predictions to improve accuracy.

(2) Resource allocation technology considering availability in the face of future demand

This technology aims to appropriately allocate resources to meet present demand while taking into account the need to allocate resources to meet the demand for new services that may be launched in the future. Although it is impossible to predict future demand with perfect accuracy, we can enhance resource availability by setting aside a sufficient amount of allocable resources (bandwidth, CPU, memory, etc.) [2].

(3) Control technology considering prediction error

Traffic engineering aims to dynamically control traffic pathways according to traffic fluctuations, but the risk exists of a drop in communications quality due to excessive control caused by prediction error or to delay-time fluctuation caused by a change in path length. This technology is aimed at achieving robust control by minimizing the effects of prediction error through sequential and stepwise control based on model predictive control technology [3].

Each of the above technologies is substantially different from conventional control technologies that respond in a reactive manner when some kind of problem (congestion, quality degradation, etc.) occurs in the network. That is to say, each can be ranked as a proactive control technology that is intended to mitigate beforehand the risk of a problem occurring.

Resource optimization enables the low-cost provision of services through efficient use of software and hardware functions across the entire network. Proactive control, meanwhile, can improve network ease-of-use for customers by mitigating exposure to problems occurring in network functions. In addition, research conducted in areas throughout network science, including disaster avoidance and quality of experience (QoE), will drive the realization of the NetroSphere concept supporting customer and operator needs.

3.1 Flexible configuration

NTT Network Service Systems Laboratories is researching and developing Multi-Service Fabric (MSF). This technology divides the network into transport functions, service functions, and control functions (controller), in contrast to a conventional network configured with specialized equipment. It enables flexible network configuration through the formation of equipment clusters that combine general-purpose products.

The MSF contributes to the transport layer that routes communications and transfers data by optical means. MSF is technology that provides extensive and flexible functions, and its name is a reference to the many ways that fabric can be used according to how it is woven and cut.

3.2 Application to networks with special requirements

This technology can even be applied to carrier networks that have special high-reliability and high-scalability requirements. MSF aims for the deployment of new networks leading to flexible operations and drastic reductions in cost. In addition, by modeling the network configuration so that modularized server and switch pools can be managed as virtual resources and dynamically controlled, NTT can clarify the requirements for switches and other functional components. In this way, we aim to encourage the participation of many and diverse vendors and promote the commoditization of network equipment.

Conventional networks have been configured using specialized equipment developed on a function-by-function basis. There has consequently been a tendency to develop advanced and large-scale devices in which high reliability and large capacity is achieved in each piece of equipment.

In recent years, however, progress has been made in increasing the speed, capacity, and performance of general-purpose products with an emphasis on datacenter applications. In addition, advances have been made in researching and developing dynamic control technologies such as NFV and SDN and in standardizing interfaces between functions.

A conventional network that tends to consist of advanced and large-scale equipment as described above lacks flexibility in terms of adding or modifying functions or capacity according to business requirements. Such a network can also drive up costs since it is not possible for low-usage portions of certain equipment to be used by other equipment. Thus, to simultaneously achieve cost reductions and flexibility in network services, there is a need not only for higher levels of flexibility and efficiency, but also for a new type of high-function and high-reliability carrier network that can transfer and transport data in a network-wide manner without being limited to conventional router functions.

Meeting this need will require new network technology that can bring existing network technology at datacenters up to carrier grade and that can make maximum use of general-purpose products.

3.3 Separated functions

MSF architecture separates the network conventionally configured with specialized equipment into three sections: (1) transport functions, (2) service functions, and (3) control functions (MSF controller), as shown in Fig. 5. The transport functions section comprises simple general-purpose equipment (general-purpose switch cluster). The service functions section is achieved on the cloud separate from the transport functions section. Finally, the control functions section (MSF controller) controls the general-purpose switch cluster. The elemental technologies supporting MSF are configuration technology, which is aimed at making maximum use of general-purpose products in a low-cost L3SW (Layer-3 switch) cluster having a high degree of freedom, and control technology implemented in the controller.

Fig. 5. MSF Architecture.

3.4 Aim and effects of MSF

Our aim in using MSF is to ensure extendibility, improve maintainability and robustness, and promptly provide low-priced services. We seek the following effects by applying a network modeled on MSF technology.

(1) Simultaneous improvements in costs and flexibility

From a transport perspective, MSF is a means of providing essential carrier functions. It will enable the application of optical network technologies at all times and facilitate the low-cost configuration and flexible control of network elements.

(2) Maintenance flexibility through hardware commoditization

MSF will facilitate the commoditization of application technologies by simplifying hardware functions and by unifying specifications. It will enable faulty hardware to be replaced by equipment made by a different vendor, thereby adding flexibility to maintenance work.

(3) Flexible scalability with no boundaries

MSF will enable any type of network to be achieved using as much common architecture as possible. It can efficiently meet the specifications of individual networks and locations with flexible configurations ranging in size from small scale to large scale. At the same time, ongoing advances in general-purpose technologies will enable MSF to provide scalability with no boundaries or limits.

(4) Modularization of network equipment in the transport layer

By modularizing and simplifying equipment and using common architecture, MSF will transform large-scale black-box transport equipment into small-scale white-box equipment. With MSF, configuring a network with only simple hardware will make it easier to visualize operating conditions at the time of a system fault and reduce operation load by enabling multiple pieces of equipment to be abstracted and managed as a single unit.

The above effects mean that a network that applies MSF will be able to reduce facility costs and power consumption, improve maintainability and operability, and provide services in a prompt and uninterrupted manner.

3.5 Plans toward expanded use of MSF

Although still in the research stage, a prototype version of MSF was constructed in 2015. The plan is to use it in experiments conducted within the NTT Group to assess and improve the feasibility of this technology. The first step in these experiments will be to cluster general-purpose switches from multiple vendors with the aim of achieving a practical technology that can replace large-scale routers. The next step will be to establish network-wide, multi-layer coordination technology for transmission equipment and other peripheral devices.

In addition, NTT is engaging in joint studies with both domestic and overseas vendors in such fields as controllers, general-purpose switches, and virtualization technology, and is participating in technical discussions with overseas carriers, all to promote the development and spread of MSF-related technology.

There are also plans to hold proof of concept (POC) demonstrations with vendors in 2015 using open technologies (related to switch/controller products) to promote the spread of MSF as a viable technology. The effective use of such POC demonstrations should help spread MSF technology to vendors and carriers in other countries.

(1) MOOS (management, orchestration, and operation system), which performs end-to-end management and control of the service components

(2) OaaS (Operations as a Service), which provides operation functions to service providers

(1) Information-gathering technology that scrambles information from operators and OSSs

(2) Analysis technology that derives useful knowledge from the gathered information

(3) Instruction technology that provides operators with that knowledge

(1) Power-supply related technology for raising the level of self-sufficiency and supplying power to network equipment

(2) Air-conditioning related technology for raising the energy efficiency of air conditioning systems for telecom buildings

(3) Technology for integrated operation of network equipment and power-supply and air-conditioning systems

(4) Technology for reducing power consumption through network architecture and network equipment designed to contribute to energy savings across the entire network

(5) Resource conservation technology for creating a green telecom infrastructure

(6) Technology for dealing with electromagnetic radiation, lightning, and other disturbances from the external environment