Beruflich Dokumente
Kultur Dokumente
Contents
What Goes on in the Mobile Packet Core? Why the Mobile Packet Core Matters More Than Ever What Operators Need in the Mobile Packet Core Where the Market Needs to Focus Going Forward
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
MME. Like the SGSN in a 2G/3G architecture, the primary purpose of the MME is
signaling so much so that many vendors tout an ability to upgrade existing SGSN platforms to MME functionality or deliver combined SGSN/MME products. And, while the MME unlike the SGSN does not actually route IP data trac to devices, to position it as simply a signaling or control plane device is to underestimate its importance. Specic functionalities supported include: mobility across 3GPP networks (e.g., 3G-to-LTE handover and SGSN selection), mobility to 3GPP2 networks (e.g., HRPD access node selection and handover), MME-to-MME mobility, idle mode device tracking, S-GW and P-GW selection, roaming support, authentication, signaling trac lawful intercept, circuit-switched fallback support for leveraging GSM/WCDMA or CDMA2000 voice.
S-GW. If the MME is the LTE analog of an SGSN in the control place, the S-GW is the LTE analog of the SGSN in the data plane. Noted earlier, the MME is not involved in the direct delivery of data trac out to end-users making the S-GW the network node that terminates the interface towards the LTE RAN. Beyond user data lawful intercept support and data packet routing/forwarding, the S-GW serves as a mobility anchor point for inter-eNodeB handover and inter-3GPP mobility (relaying trac between 2G/3G systems and the P-GW). At the same time, it buers LTE RAN idle mode downlink packets, performs transport layer packet buering and provides user-level session granularity accounting for inter-operator charging purposes. P-GW. Where the S-GW terminates the interface towards the LTE RAN, the P-GW is the
node terminating the interface towards the packet data network (PDN) the LTE analog to the GGSN. More appropriately, it terminates the interface towards a single PDN; if a user device is accessing multiple PDNs, that device may be associated with more than one P-GW. As the gateway out to external data networks, the P-GW is responsible for a number of critical functions: policy enforcement, DHCP and IP address allocation for subscriber devices, charging support, packet screening and ltering thanks to deep packet inspection capabilities. Further, for handovers between 3GPP and non-3GPP networks (CDMA2000, WiMAX) the P-GW acts as the mobility anchor point.
PCRF. Policy control is an integral part of 3GPP standards for 3G as well as LTE. In LTE
networks, however, it takes on new signicance: again, all-IP operations make the treatment of dynamic, application-specic QoS more critical than in the 3G space where circuit connectivity is still an option. Implied by the name, the PCRF represents a decision point for allowing or disallowing QoS requests based on application requirements, user subscriptions or user requests with enforcement being carried out in the P-GW. Of course, policy and charging are tightly tied: with application or user-specic policy representing something an operator could potentially earn added revenues from. Noted earlier, other technologies outside the 3GPP family have their own packet core architectures. WiMAX relies on an Access Service Network gateway to support connectivity and inter-cell as well as inter-operator mobility. CDMA2000 networks rely on a packet data serving node delivering the functionality of a SGSN and GGSN. Ultimately, however, with CDMA operators moving to LTE and operators with unpaired spectrum resources investigating TD-LTE, EPC looks to be a near-uni-
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
nobody. As the mobile Internet moves from interesting concept to everyday reality the impact on the core is nothing short of obvious. Consider the projections Cisco has made around the growth in mobile data services. Starting from just over 90 TB of mobile data trac per month, the companys forecasters (based on insight into the trac running over their own gear) predict a 108% compound annual growth in mobile data usage topping out at 3.565 TB of data trac just ve years into the future! Driven by new technologies such as HSPA, HSPA+ and LTE, its clear that the packet core deployments and solutions of several years ago just wont be able to support this growth. Predictions (or extrapolations) of growth in mobile broadband trac, however, only tell one part of the story. From the service provider perspective, mobile trac growth is about more than just the prospect of 3.6 Exabytes of monthly mobile data in 2014. Its about more mobile broadband users. Its about multiple sessions per user. Its about more powerful devices which make mobile data services more accessible and meaningful. Its about new applications that make these data services an integral part of our lives in the same way that mobile voice services evolved from a luxury into a necessity. Its about the simple fact that these applications are all built on IP. And, what does this all mean? It means that the mobile packet core, once a simple necessity for supporting avant garde mobile data users, has developed into a critical network asset a tool for helping operators deliver table stakes data services, better monetize those services and grapple with the undeniable realities of evolving mobile broadband usage, including: More Data Per User. Beyond the simple fact that more people are embracing mobile data services, the throughput being delivered to those users is trending upwards as well and looks to continue doing so through the medium term. Just a few years ago, 7.2 Mbps HSPA was an impressive feat. Today, operators are rolling out 28 Mbps and 42 Mbps oers. Verizon Wireless recent LTE network testing in Boston and Seattle suggests peak downloads up to
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
Solution Completeness With the exception of direct tunnel architectures, operators need complete packet core solutions not just products. An SGSN without a GGSN doesnt do much good. An MME without a P-GW and S-GW is useless. This doesnt mean that vendors who only address the signaling plane or the data plane have no value; since the interfaces between these products are standardized, it is completely possible to build multi-vendor packet core networks. Instead, it simply means that operators need to think about the core as an end-to-end proposition in the same way they think about their end-to-end LTE network.
Single-Vendor Solutions. Vendors promising a complete mobile packet core or EPC oer need to deliver on all components including SGSN/GGSN or MME/P-GW/S-GW and PCRF, pointing to the proven interworking between, and common management across, them. Multi-Vendor Interoperability. Where a vendor produces only one component of the 2G/3G or LTE packet core, operators must know that its products seamlessly t into a broader solution. In other words, the vendor must lead with proven interoperability ideally
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
Capacity & Scalability Speeds and feeds represent classic operator buying criteria across the telecom infrastructure landscape making them some of the most commonly quoted performance metrics by vendors. At its most basic level, the concept that more is better (like megapixels in a camera or horsepower in a car) holds true. Overall, capacity signals how well a platform can actually support current network demands along with trac growth going forward. In the context of LTE promising to push more data out to more users, the obsession with capacity in the EPC is completely logical, extending to:
IP Bearers. The LTE equivalent of 2G/3G PDP contexts, IP bearers consist of data ows including user session information. Subscribers. While a given P-GW or S-GW can serve a nearly innite number of users over the course of its operations, the number of simultaneously attached users it can support represents its maximum active user capacity at a given point in time. Throughput. P-GW/S-GW throughput directly equates to the amount of bandwidth available to support data-intensive subscriber sessions. With LTE operators already promising capacity upwards of 5 Mbps on the downlink, throughput gures need to be evaluated in the context of subscriber capacity i.e., how much throughput is available per subscriber under full utilization? Activations. Again, high-quality data services require sucient bandwidth in the packet core. At the same time, users must be quickly activated on (attached) or deactivated (detached) o of gateways in order to keep service latency in check and avoid subscriber capacity exhaust. RAN Fan Out. Increasingly, operators are considering small cell architectures as a way to deliver robust LTE RAN capacity and coverage especially when launching in high-frequency spectrum bands. This implies an S-GW that can support a large number of eNodeBs.
Application Support
Our characterization of the mobile packet core and EPC as conduits for trac between the RAN and data networks was an admitted over-simplication. More than just dumb pipes, the packet core is responsible for providing trac intelligence (i.e., what trac is running across the network) and then leveraging that intelligence to deliver a better user experience and help operators better monetize their services, thanks to: Security Scalability. Mobile data gateways obviously need to support security measures like rewalls (personal, stateful) and IPSEC generation. How well they scale this support how much security they can oer will determine how secure an operators user base truly is. Deep Packet Inspection (DPI) Capabilities. DPI has become table stakes in the packet core as a tool for understanding what types of data are running across an operators networks, setting the stage for dierentiated billing, and other monetization plans. How much insight it can provide i.e., Layer 7 visibility determines how well a solution can execute on these promises. Trac Analytics. DPI may provide insight into session content, but solutions for visualizing or reporting on this trac are critical for helping operators make sound decisions around how to manage their trac: which users or applications are clogging the network, when congestion is occurring, when usage can be encouraged, etc.
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
Deployment Flexibility Compared with the massive amounts of user trac running across a packet platform or its ability to identify (and manage) specic trac types, considerations around actually operating the platform may seem mundane, at best. They are not - beyond the issues of site acquisition and platform sizing, simple or familiar management tools help to ensure service reliability while promising to keep management-related OpEx in check.
Platform Familiarity. Even where a new gateway oer promises incredible performance leaps, consideration must be given to how simply an operator can integrate the platform into its network. How much new training will be required? Will new stocking and sparing processes be needed? Gateways that have been well deployed or built on well-deployed platforms clearly provide an advantage. Proven Operations. It is unfair to suggest that all vendors exaggerate the performance of their solutions. That said, it happens particularly with new solutions where capabilities may still be in development. Where performance claims have been backed by real world operations, this worry is somewhat mitigated. In the absence of (or run-up to) widespread deployments, lab testing can be a proxy. Combined Operations. Having deployed mobile packet core gateways for their 2G/3G networks, EPC solutions can be deployed as an overlay. Vendor oers supporting EPC functionality alone suggest an interest in the strategy. However, if solutions can support both 2G/3G and LTE networks, the CapEx, OpEx and management eciencies are clear. Sizing. Platform sizing relates directly to how expensive (or inexpensive) a mobile packet core solution is to site. Perhaps more importantly, sizing becomes an issue where vendors claim impressive scalability thanks to massive gateways which might occupy several times the physical space of competing products. Redundancy. Any network asset that impacts the service availability and quality of millions of users will be deployed with redundancy. In the mobile packet core, how redundancy is achieved directly impacts how a gateway is deployed, how many need to be deployed (2N vs. N+1, etc.), and how meaningful capacity gures are when judged from a per gateway or per solution perspective.
2010 Current Analysis Inc. All rights reserved. For more information, please call +1 703 404 9200, toll-free +1 877 787 8947 Europe +33 (0) 1 41 14 83 15. Or visit our Web site: www.currentanalysis.com
10