The next wave of SLAs

Dec 28, 2001 12:00 PM

Service providers are offering more and more services to fulfill the need of enterprise wide-area connectivity at higher bandwidth than what it is offered today -- services like native gigabit Ethernet, native storage connectivity or simply a Sonet/SDH transport pipe. To differentiate themselves, some service providers are currently associating service level agreements (SLAs) to these services. SLA metrics are provided in addition to service connectivity, which affects the architecture and the choice of transport technology in the design of the optical-based service transport networks.

The first step in managing SLAs is to provide accurate measurement of service availability. This is the most critical parameter because it reflects the end users' perception of the quality of the services being offered to them. Limiting our scope to only the connectivity aspects of the service, the service transport quality can be measured in bit error rate (BER), packet loss or other equivalent measurements. In addition, planned outages for maintenance purposes in defined and scheduled maintenance windows become the two major elements of measuring service availability.

Another important SLA parameter for a transport service is the measurement of service provisioning and service restoration time. Other parameters include service throughput and service latency. 

Can all these SLA parameters be applied to a wavelength-based service?

When offering a wavelength-based service, traffic is mapped natively onto individual wavelengths. Access to the content or format of the traffic is an option. With native gigabit Ethernet services, the "handoff" between the service provider and the end user is a simple fiber connection. The service is logically and physically terminated directly onto the end user's IP router or L2/L3 switch. The service is then transported natively across an individual wavelength over the metro wavelength division multiplexing (WDM) network again to be terminated on another IP router or L2/L3 switch. A storage service such as native Fibre Channel transport across the metro network is deployed in a similar fashion. The service is directly terminated into a Fibre Channel switch at the end user side. The storage traffic is then mapped onto another optical wavelength, which is transported over the WDM metro network. 

In both cases, the service provider cannot view the format of the service. IP packets and storage data blocks are being transported transparently across the optical WDM network.  This transparency allows the delivery of the service to be more cost-effective but at the same time makes the measurements of SLA metrics difficult to implement. Without optical-based monitoring capabilities, measurement of service availability in transparent service transport system is reduced to measuring the physical connectivity of the service (i.e., the service is available if the optical signal is on, and the service is not available when the optical signal is off).

Service providers can choose to map every service onto Sonet/SDH because such transport technology has intrinsic parameters, which measure the availability of the service through extensive performance monitoring capabilities. Sonet/SDH OAM&P parameters are used to measure the availability and the quality of the transport service.  Parameters such as code violation, errored second and severely errored second are used to provide SLA metrics with which service providers can build SLA reports for their Sonet-based transport services. Parity monitoring (B1/B2 bytes) is used to calculate whether the encapsulated data transmission from one Sonet terminal to another contained any errors.

Different SLA metrics are used for packet-based services. In addition to the guaranteed network/service availability of 99.99 percent, service providers are providing latency and packet loss metrics, two critical performance parameters affecting the aggregate throughput of IP-based services.  

Not all services can be mapped onto Sonet/SDH

Minimizing the mapping function of services onto Sonet/SDH can lower the cost of such solutions. This method is currently being referred to as "thin Sonet." With such a mapping method, Sonet/SDH frames and overhead information are added to the service payload. Extensive Sonet/SDH overhead support is forsaken to reduce complexity. But no matter how "thin" the mapping is, cost and a certain degree of complexity are added. In addition, most of the high-bandwidth native transport services such as gigabit Ethernet, Fibre Channel and FICON cannot be gracefully or optimally mapped onto Sonet/SDH frames. The overhead and the complexity make such solutions unattractive if bandwidth scalability is required. Direct mapping of services onto optical wavelength is needed to deploy efficient, cost-effective, wavelength-based service delivery systems.  

Delivering native gigabit transport services on individual optical wavelengths requires the highest level of service availability. High bandwidth means more users are being connected and/or more information is being transported per service. Nevertheless, service providers are offering end users a range of choices with different SLA flavors for their gigabit transport services. Differentiation between flavors is basically determined by the degree of resiliency and service uptime. Clearly, the higher the resiliency, the more costly the service is. With SLA metrics being provided on a per-service basis, flexibility in the resiliency aspect of WDM metro network architecture becomes essential. Because native gigabit services are mapped onto individual optical wavelengths, resiliency must be provided on wavelength basis. In addition, service monitoring and SLA measurements must be implemented on each individual wavelength. Service providers must offer SLA metrics for each wavelength within their metro access and metro core WDM networks. 

Because resiliency is crucial in maximizing service uptime and network availability, service providers are deploying resilient WDM systems in the metro access and the metro core networks by using Sonet/SDH-based topology concepts. Optical unidirectional path switched ring (UPSR) like solutions are being deployed in metro access and metro core WDM networks. Optical UPSR-like rings enable the service providers to offer the following service resiliency parameters, which directly affect and shape SLA implementation:

• Per-wavelength path protection only

• Per-wavelength service interface protection only

• Total wavelength protection (service interface and path protection)

Clearly, service providers can provide "no protection" as an additional option with associated SLA guarantee of service availability in the order of 99.95 percent or 99.96 percent.

This option is attractive to end users who would implement their own connectivity protection scheme at Layer 2/3 or at the Sonet/SDH level. Under this scenario, the end user would subscribe to two unprotected wavelength services with disparate routes. The connecting devices at the customer premises would perform the switching in case of fiber transmission failure. Such an option is widely used in IP-based networks and services.

The path-protection-only option is designed to allow end users to take advantage of the optical ring infrastructure, which offers two separate paths between any two end-points of the network. In this case, service providers are providing a protected wavelength along with the working wavelength. This is referred to as 1-to-1 service path protection, which maximizes service uptime and increases service availability.  In case of a fiber cut, the transport of the service is switched from the working wavelength to the protected wavelength within 50 msec. With such a protection scheme, the SLA availability metric can be defined as high as 99.95 percent.  With such options, end users will expect an average bit-error rate of 10-10 on their service.

Providing per-wavelength service interface protection only may not be very attractive to most end-users. Such a scheme is designed to provide 1-to-1 hardware equipment protection only and does not support path or route protection. The combination of both path and service interface protection would constitute total wavelength/service protection.  With such a protection scheme, SLA availability metric could reach 99.995 percent or 99.996 percent or even higher depending on network downtime due to failure other than transmission and transport. With the total wavelength protection, maximum resiliency is offered for both hardware and software elements of any given wavelength-based service.  End users have double the hardware in addition to the dual path or service routes. 

Resiliency in network topology is principal in delivering services with optimum uptime. The other important element in offering SLA-based wavelength services is monitoring. Real-time performance monitoring and management is essential to the implementation of wavelength-based SLA services. Service providers have to rely on service monitoring techniques deployed at the transport layer rather than at higher network layers. Such techniques need to provide:

• Monitoring of the raw bit stream (bit-error rate) in real time at multi-gigabit rates with high accuracy

• Monitoring the optical signal transmission while in-service

• Detecting degradation and pro-actively responding to ensure SLA

• Performing the above regardless of protocol (IP, ATM, Gigabit Ethernet, Fibre Channel, and SONET/SDH)

Such monitoring schemes should be implemented non-intrusively and should be bit-rate independent to be scalable across the whole service list of service providers.  To be deployed successfully, such monitoring techniques must not introduce latency and should not require bandwidth overhead.

Service providers are beginning to use such non-intrusive monitoring capabilities in their WDM metro core and metro access networks to determine physical level SLA metrics that measures the optical signal link integrity and generates QoS metrics on live services.    Such capability enables service providers to create and monitor SLAs based on optical bit error rate on any wavelength-based service. With these metrics, service providers can create SLAs per customer, per service, per link or per wavelength. Periodic reports can be generated for billing and trouble ticket usage. 

To provide a complete set of comprehensive SLA metrics upon which service providers can build an SLA report for their gigabit wavelength-based services, service performance metrics addressing service availability and service performance must be provided at the optical transport level.  Such parameters must be provided on a per-service (a group of connections per wavelengths) and on a per-connection basis (for each individual service path or wavelength).

Once the above parameters are implemented, service providers can define and negotiate SLAs with end users based on these performance metrics. The guarantees in an SLA could be further defined as a combination of one or more measurements to which constraints are applied. User configurable threshold level of any performance metric could be specified with a variety of time intervals (i.e. time of day, day of week over which computation is performed). Critical alarms measurements could be included as part of the SLA metrics. Alarm computation policy options could be offered to enhance the service offering. 

Service providers will use wavelength-based SLA services as a service differentiator. Providing monitoring capabilities inherent to the optical transport is the critical element in making wavelength-based SLA successful. As WDM networks are deployed, monitoring at the optical transport layer becomes more of the norm than the exception to the rule. Once deployed in the network, it will be the basis for measuring service uptime and availability, two paramount parameters in the generation of optical-based SLA services.

Paul Zalloua is Assistant VP of Product Marketing for LuxN Inc.

Visit LuxN online.