Choose the right formula for compute-intensive HSS systems: AdvancedTCA platforms versus 1U rack-mount servers

AdvancedTCA allows a modular approach to IMS network performance issues.

Based on current overall network growth, Telecom Equipment Manufacturers (TEMs) and Network Equipment Providers (NEPs) are being challenged to launch new communications platforms that provide optimal scalability, flexibility, manageability, and high computing performance. With the expected increase in new subscribers who need to be supported across the network, TEMs and NEPs are focusing on COTS platforms that meet the requirements of a carrier-grade environment. With these new challenges, it is important to take a look at the different applications in the network to thoroughly evaluate the best hardware platform approach. Comparing the scalability, I/O functionality, interconnectivity, and overall performance gives TEMs and NEPs a solid basis from which to choose the best cost-effective solution for their applications. This article compares the traditional 1U PCI platform with a more modular AdvancedTCA-based approach to a Home Subscriber Server (HSS)-based platform.

Emerging next-generation network architecture: IMS

Carriers and service providers are focused on improving their Average Revenue Per User (ARPU) by deploying new services quickly and at a price that is acceptable to subscribers. IP Multimedia Subsystems (IMS) enable service providers to offer differentiated services to their subscriber base, integrating voice, video, text, and content as part of their service mix regardless of the access technology.

The IMS framework provides the telecom industry with a modular, standards-based IP service delivery infrastructure. More than simply replacing existing circuit services with similar services delivered over IP, IMS allows service providers to capitalize on faster and more cost-effective deployment of new and differentiated multimedia communication services.

While these new technologies promise new sources of revenue and enhanced customer loyalty, they also pose implemen- tation challenges. One of the crucial network elements within an IMS framework is the manageability of the subscriber base. Today an increasing number of users subscribe to multiple services such as multimedia mobile phones, VoIP, or IPTV at home. Service providers must manage “duplicated” subscriber databases for each particular service. Consequently, managing a converging subscriber base for such network functions as billing across multiple different technologies presents a tremendous challenge. Any downtime of this network element can be costly to the carrier or service provider, not to mention the heavy risk of losing subscribers over the long term.

The HSS and Home Location Register (HLR) are key elements in the IMS network required to manage the growing number of subscribers. These platforms that manage millions of customers’ data need to be highly reliable with carrier-grade, High Availability (HA) functionality. Core requirements for the HSS computing platform are:

  • High compute density for application processing
  • Scalable performance and high density
  • Fast read access to the subscriber data
  • Large storage and memory bandwidth
  • Ability to consolidate multiple IMS elements and subscriber data within a single platform with a modular approach as a converged services platform

HSS plays network backbone role

HSS provides the important network backbone of the IMS architecture because it eases innovation and addresses a range of service-logic interaction issues that have plagued previous service architectures. HSS offers open access to service-related data for each subscriber, supporting the sharing of data among multiple services.

HSS is essentially the master user database that supports IMS network entities that handle calls and sessions. It contains user profiles, performs authentication and authorization of the user, and can provide information about the physical location of the user. It plays a key role in provisioning, service creation, enabling subscriber data, and managing control, roaming, and interconnection. The Application Server (AS), which hosts and executes serv- ices in the IMS environment, communicates with the HSS. Call State Control Function (CSCF) servers also communicate with the HSS, which must be able to:

  • Support IMS-level authentication and authorization
  • Maintain IMS subscriber profile(s) independent of access types
  • Maintain service-specific data
  • Keep track of currently assigned CSCF
  • Support CSCF and AS access

HLR/HSS hardware platform design choices

HSS design must enable the HSS to be deployed cost-effectively and expand readily to meet future subscriber levels. For example, the trend to include more information as part of the data record for each subscriber will drive up memory requirements.

From a technical perspective, the key metrics to understand when developing the IMS HSS functions are the number of completed transactions required per second and the size of the subscriber base.

HSS architecture typically comprises multiple processor units with extensive memory requirements. With the introduction of multicore processors from Intel, developers can scale performance while optimizing power consumption in the overall system design.

To meet these requirements, the hardware platform needs to provide scalability, flexibility, and long-term cost efficiency. In comparing the traditional 1U platform with a more modular approach to the HSS using AdvancedTCA, consider the following:

  • High compute performance
  • High memory and storage capability
  • I/O functionality to connect to multiple networks (such as SS7 and Ethernet)
  • Carrier-grade profile and redundancy

22
Figure 2

AdvancedTCA: High density with scalability on a smaller footprint

Table 1 illustrates a comparison of a sample AdvancedTCA-based platform and a 1U PCI rack-mounted platform for HSS applications. This configuration assumes an increasing subscriber growth from 1 million (M) subscribers to 18M subscribers per platform.

Footprint

AdvancedTCA-based platforms provide a high density compute platform with scalability on a smaller footprint. TEMs can easily add boards to a chassis to accommodate more traffic and to support new services. This flexibility enables an increase in subscribers and the database without having a major impact on rack space in the central office environment. To meet the same capacity for 18M subscribers, the 1U PCI approach will occupy two bays or cabinets with a total footprint of 28 square feet compared to the same capacity occupying one bay or cabinet with a footprint of 14 square feet (see Figure 1).

Cooling

While the power consumption per square foot is higher than the 1U PCI platform, AdvancedTCA offers better airflow and cooling distribution using a push-pull cooling system. Each 1U PCI platform includes up to nine cooling fans and is dedicated to that CPU in the 1U server only, which translates into higher overhead. Therefore the airflow is not optimized in the overall system design. In the case of a fan failure, AdvancedTCA platforms enable easier access to replace the fans, which are field replaceable units. The 1U platform also does not include any integrated centralized switching or management capabilities; therefore the total does not include any switching or system management functionality.

Scalability

AdvancedTCA architecture lends itself well to scalability. As AdvancedTCA supports both a Base and Fabric interface into the backplane and the separation of data and control planes, TEMs can also scale up their designs from 1 GbE to 10 GbE to accommodate new users and usage requirements. AdvancedTCA based systems are designed for easy upgradeability to new processing blades, switch fabrics, and I/O modules using Advanced Mezzanine Cards (such as SS7).

Since AdvancedTCA and AdvancedMCs offer hot swap capabilities and HA functionality, growing a subscriber base by using additional CPU blades within the system with low maintenance or upgrade downtime is crucial for the HSS plat-form. It also allows for the up and down scaling of the platform depending on the subscriber increase or decrease. When consolidating multiple services, minimizing multiple duplicate databases requires this flexibility.

Interconnectivity

The AdvancedTCA platform offers a number of interconnectivity features including:

  • Separation between the data plane and the control plane
  • Direct uplink in a multichassis environment with up to l0 GbE
  • Node to node base and fabric connectivity with HA carrier-grade Ethernet functionality
  • Easy upgrade path to l0 GbE for increased bandwidth

By contrast, the 1U PCI platform requires:

  • Additional external interconnectivity between 1U rack-mount servers, which increases the number of points of failures and the operating and maintenance costs
  • Additional switching equipment for carrier-grade Ethernet connectivity
  • Additional manage-ment (Simple Network Management Protocol (SNMP) for HA redundancy

Power

The 1U-based platform also utilizes integrated AC power supply units, usually the standard for the data center and enterprise environments. AC will increase the provider’s OPEX in the long term, since the power consumption per rack is significantly higher and requires additional support equipment such as an Uninterruptible Power Supply (UPS) and an AC to DC converter per rack space, and ultimately per CPU. These factors, along with the requirements for carrier-grade functionality, HA support, and a more favorable long-term OPEX, further reduce the need to consider the 1U solution for HSS deployments in Central Office networks.

Manageability

AdvancedTCA chassis-level management is based on the industry-standard Intelligent Platform Management Interface (IPMI). IPMI makes possible interoperability among chassis and blowers, compute nodes, switches, and management solutions from multiple vendors, enabling these elements to work together seamlessly. The resulting successful inter-blade communication for applications, including failover and clustering, provides full manageability on a per shelf and per bay basis.

One of the primary goals of the AdvancedTCA specification is to ensure a very low Mean Time to Repair (MTTR) and high Mean Time Between Failures (MTBF). There are no active components in the backplane – which can translate to expensive repairs in the field. In addition, AdvancedTCA specifies that all components can be hot swappable in the field, and the architecture provides redundancy at all levels. This hot swap capability translates to less downtime for field replacement on a per CPU blade basis, as compared to 1U PCI.

21
Figure 1

Summary

AdvancedTCA provides an ideal platform and meets the technical requirements for deploying flexible and scalable IMS HSS applications. While other platform architectures such as 1U rack-mount servers may with considerable difficulty be able to deliver HSS capabilities, AdvancedTCA offers a much more advantageous solution by providing a smaller footprint, improved cooling and thermal characteristics, better scalability, enhanced interconnectivity, and easier manageability. This all contributes to making AdvancedTCA a lower cost approach in terms of CAPEX and OPEX for long-term subscriber growth.

1U PCI rack-mount servers are not considered to be the optimal hardware platform for HSS applications in a Central Office. With its limitation for carrier-grade flexibility and HA functionality, the 1U PCI platform is best suited for a full range of other enterprise and Internet infrastructure applications within a data center environment that does not require any significant flexibility and HA functionality around it. It might also be a good entrance point for other enterprise applications with lower price points.

The use of AdvancedTCA-based platforms enables TEMs to shift away from hardware development and refocus their efforts on delivering value-added software and services. As a result, they can realize lower development costs and faster time-to-market, enabling increased flexibility with each product design through the use and reuse of standardized off-the-shelf components. Equally significant are the advantages that can be realized by the service provider. With a common platform and a flexible architecture that maximizes limited floor space, service providers can rapidly and flexibly implement a broad range of feature-rich solutions for faster profitability.

Sven Freudenfeld is responsible for North American Business Development for the Kontron AG line of AdvancedTCA, AMC, MicroTCA, and Pre-Integrated Solutions. Sven possesses more than 15 years of experience with voice, data, and wireless communications, having worked extensively with Nortel Networks in Systems Engineering, Sanmina-SCI in Test Engineering, and Deutsche Telekom in Network Engineering. Sven holds an electrical engineering degree from Germany, and is also Chair of the CP-TA marketing workgroup focusing on the interoperability of COTS standard building blocks.