Soapstone Networks - Avici's new open control plane solution business

Avici Systems has established a new business unit, named Soapstone Networks, to market an open control plane solution designed to simplify the operation of complex, heterogeneous networks.

Soapstone's software solution is designed to be used by carriers to enable the delivery of high value services over next generation networks, and by equipment vendors to add value to low-cost switching products.

Soapstone Networks is applying Avici's expertise in routing to a new area of the market. The software product for next generation networks creates an abstraction layer as a bridge, or glue, between the network and applications running over it.

The solution serves to hide the complexity of the technology and structure of the transport domain and support the resource side via abstractions such as TMF and IPsphere.

This effort is being done largely outside the IETF which has traditionally taken a protocol approach to solving problems.

The new solution is designed to present a common control framework for underlying technologies, including IP, Ethernet and optical, and enable migration between the different network technologies.

From the original circuit-switched voice network, as new services have been developed new networks have been deployed to carry them, so that there are now separate Frame Relay, ATM and IP networks. Over time, the relative importance of the different networks has shifted.

Today, data traffic is increasing dramatically and IP technology is maturing. Carriers are assessing their systems and attempting to predict what services the network will need to support and the type of competition they will face in the future.

The overall conclusion is that the way networks are architected needs to change to allow operators to address the market effectively and compete in the new environment.

Carriers are basically seeking technology agnostic services and applications. For example, with VoIP the carrier simply wants the IP call to go through, whether it is carried over an MPLS-enabled, Ethernet or native IP link.

Carriers also want to be able to control multiple services, via IMS for example, and to be able to manage future services via the same architecture.

In addition, consolidation is taking place at the transport layer utilising technologies such as IP/MPLS, Ethernet PBT and all-optical networks.

At the other end of the scale, network access methods are diversifying with technologies such as WiFi, RAN, Ethernet and CPE devices from companies such Microsoft.

Carriers are implementing centralised provisioning processes such as AAA and subscriber and policy management based on SLAs. These processes are based on building blocks that can be reused or reconfigured to meet changing needs.

Currently there are many services being delivered over the network, such as voice, DSL, ATM, VPN, together with a simultaneous explosion in access methods. The result is that numerous services need to be delivered to an increasing range of access devices.

In response, carriers are seeking to consolidate the number of platforms in the 'middle' of the network - the transport domain. Convergence is effectively being driven by pressure from the top - the services offered, and bottom - access devices available.

To run correctly on NGNs, services require clear APIs that can instruct the transport and control layers of the network how to treat different traffic types.

Carriers also wish to leverage the transport network, in which they have invested heavily, to increase revenue opportunities. To achieve this they want to couple the transport network to the services it delivers to create a hybrid service offering that is differentiated from the competition.

A further issue for carriers is that services compete with each other for resources in the transport layer. This necessitates service intelligence to enable effective management of different services as they move over the transport network.

With network transformation to IP comes the issue of 'touch'. Carriers may be running Frame Relay and ATM networks that support the high end, 'high touch' services valued by business customers.

While the carriers wish to migrate from expensive Frame Relay and ATM networks to lower cost IP networks, they cannot afford to lose the ability to deliver legacy 'high touch' offerings.

Today there is a lack of APIs enabling service elements to specify attributes such as bandwidth, failure recovery procedures and delay characteristics in the transport domain. For applications such as gaming, the ability to define jitter and packet loss characteristics is essential.

The bottom line is that new services require absolute parameters on a service-by-service basis, rather than the traditional, generalised bronze, gold and platinum service levels.

With diversification of services, there must be somewhere on the network where service arbitration can occur. In addition, high value services need to be able to communicate their requirements to the network regarding procedures such as failure recovery so that the system restores the most important services first.

Ultimately, carriers wish to link business needs with the operation of the network.

At the same time, the network continues to evolve - from IP to IP over MPLS and now to lower cost technologies such as PBT and Ethernet.

A further complicating factor is the number of boxes in the network - from multiple vendors, of different ages and with different capabilities. There is a need to reconcile the technologies and capabilities comprising the network in order that services can make optimum use of it.

There is a fundamental requirement for centralised coordination of all of these disparate elements.

Soapstone aims to deliver a network abstraction layer that decouples the services from the network infrastructure. The company is providing the glue between the IT-centric management domain and the network-centric service delivery layer.

Soapstone provides mediation between the service and the transport network and OSS systems, while negotiating the complexity of the underlying network.

Avici has utilised its expertise in carrier transport network elements and its active participation in industry consortiums, developing service creation and abstraction technologies, to develop the Soapstone solution.

Abstraction is important as it allows emulation of Frame Relay and ATM features - the 'high touch' functionality that customers value - over NGNs.

Soapstone provides service-facing APIs that are layered over virtual resource abstraction. Resource abstraction describes abstract behaviours of resources in the network to enable the mapping of service requirements to available resources.

As an example, a low resource-intensive application could be carried via IP, a more demanding application could use MPLS. In the access network, Ethernet or PBT could be used, or for a high bandwidth service an all-optical link could be selected where available.

As previously mentioned, carriers wish to reduce the cost of 'high touch' ATM and Frame Relay-based services without losing the features that customers value, such as maintenance, billing and SLAs.

IP is a highly resilient technology - if a link is lost traffic is re-routed. The downside is that resiliency means that the service provider loses touch with the traffic and does not always know where a particular packet is or exactly how it will behave. For this reason, the end customers are reluctant to move from legacy ATM to IP as the 'high touch' features will be lost.

A key issue for carriers is being able to inform a customer of the status of their service at the first call to a support centre if a fault occurs, as is the case with a Frame Relay or ATM-based service. The carrier does not want to have to escalate a call because first-line support staff cannot tell the customer exactly what is happening.

To address this issue, protocols such as MPLS were introduced as a layer below the IP domain. However, in the network, MPLS is actually implemented both above and below IP. Therefore an IP control plane is required to discover which network elements are active; MPLS is then run on top of this.

However, despite such enhancements, there remains a degree of uncertainty regarding where each packet is on the network at any time. This issue is currently being addressed using technology such as PCE (Path Computation Element), but is yet to be resolved.

When developing a scalable router, Avici separated the data plane from the control plane. In doing so it noted that 95% of the cost of a router lies in the data plane and that 95% of features are associated with the control plane.

A separate control plane based on IP principles but with enhanced policy controls can be applied not only to IP-based services, but also to services at lower layers such as Ethernet PBT, hybrid Ethernet / optical and optical.

Provider Backbone Transport (PBT) is a technology that can be leveraged to deliver high value services at reduced cost compared to existing solutions.

Removing the control plane element from an Ethernet switch and putting it in the OSS enables commoditisation of the hardware - the value then resides in the software providing the control plane rather than in the hardware. This is accomplished by placing a simple API on top of the switch.

To make PBT a basis for service elements it needs to be a part of an open control plane - carriers do not want yet another control plane that is solely for PBT.

Countering this shift is the desire of equipment vendors to retain the control plane within their systems as a barrier to low-cost competitors.

Avici, as one of the very few developers of high reliability control planes for IP services, wishes to facilitate the adoption of PBT by offering an open control plane separate from its hardware.

As carriers have already adopted MPLS and invested in equipment to support it, Soapstone aims to offer carriers a choice of MPLS or PBT. By providing an abstraction layer that overlies both technologies, services can be mapped to either MPLS or PBT as desired.

Additionally, Soapstone makes it possible to migrate to PBT from MPLS without a flag day scenario to take advantage of PBT as a high-touch alternative.

PBT behaves as a policy-less data plane. The bottom line as regards policies is the carrier's business, therefore business policy should define network policy. Currently there exists no abstraction that enables this flow of information.

Soapstone sits between the business and network layers where it translates business policy into network policy, so enabling this flow of information.

An example of a higher layer application service that could be handled by Soapstone is IMS. At the highest level, IMS simply establishes a session between two gateways over any available network technology. A key factor is the SLA, therefore the application is resource aware.

Soapstone can provide an API to an IMS RACF, choose the optimal transport elements, provision the network, and monitor behaviour.

By managing virtual resources and providing views into alternative management frameworks, Soapstone can coordinate virtual resources between high-level frameworks.

Soapstone is intended to leverage technologies such as PBT to enable carriers to reduce costs and allow the use of next generation OSSs. It also simplifies the delivery of MPLS-based services.

Other router vendors are currently unwilling to take the cost out of their hardware by separating the control plane element to facilitate simpler networks, as this is the foundation of their business.