LABORATORY 117 Charging for Mobile All-IP Telecommunications Yi-Bing

LABORATORY 117 Charging for Mobile All-IP Telecommunications Yi-Bing Lin Chair Professor and the Dean College of Computer Science National Chiao Tung University Hsinchu, Taiwan liny@csie. nctu. edu. tw 2018/3/15 1

LABORATORY 117 Outline IP Multimedia core network Subsystem (IMS) IP-based Online Charging System (OCS) Design of IMS Prepaid Application Server for SIP-based Services Modeling OCS Credit Reservation Procedure Reducing OCS Re-authorization Cost Future Research Directions 2018/3/15 2

LABORATORY 117 Motivation Internet environment encourages global usage with flatrate tariffs and low entry costs. A major problem of the “flat-rate” tariffs is that such business model cannot justify the expensive equipment/operation investments of mobile services. Introduction of the 3 G mobile system has further driven the Internet into new markets to support mobile users. To have cooperation with service/content provider, creating a killer environment has now become a key point in terms of competition. 2018/3/15 3

LABORATORY 117 UMTS/IMS Online Charging GSM supports circuit-switched telephony services Mobile Switching Center (MSC) UMTS supports real-time IP multimedia services General Packet Radio Service (GPRS) IP Multimedia core network System (IMS) Charging Management Evolution SS 7 -based (Circuit Switched) Single Session 2018/3/15 IP-based (Packet Switched) Multi-Session 4

LABORATORY 117 UMTS/IMS Architecture c d a 2018/3/15 b 5

LABORATORY 117 UMTS Online Charging System CCR CCA ABMF: Account Balance Management Function CGF: Charging Gateway Function RF: Rating Function 2018/3/15 6

LABORATORY 117 Support of IP-based Charging Mobile Operator Service/Content Provider Users 2018/3/15 Prepaid and Postpaid Dynamic Pricing policy Single Billing Transparent of the Price 7

LABORATORY 117 Research Topics Diameter Credit Control Protocol [1] CCR: Credit Control Request CCA: Credit Control Answer [2] [3] [1] Multiple Service Charging between the Prepaid Application Server and the OCS [2] Online Credit Reservation between the CSCF and the OCS [3] Credit Re-authorization between the GGSN and the OCS 2018/3/15 8

LABORATORY 117 [1] Design of IMS Prepaid Application Server for SIP-based Services S. -I. Sou, Y. -B. Lin, Q. Wu and J. -Y. Jeng, Modeling of Prepaid Mechanism of Vo. IP and Messaging Services. IEEE Transactions on Vehicular Technology, 56(3): 1434 -1441, 2007. S. -I. Sou, Q. Wu, Y. -B. Lin and C. -H. Yeh, Prepaid Mechanism of Vo. IP and Messaging Services, IEEE International Conference on Information Technology Research and Education (ITRE), Hsinchu, Taiwan, 2005. S. -I. Sou, Q. Wu, Y. -B. Lin and W. -E. Chen, SIP-based Vo. IP Prepaid System on NTP Vo. IP Platform, Proceedings of the Taiwan Academic Network Conference (TANET), Taitung, Taiwan, 2004. 2018/3/15 9

LABORATORY 117 SIP-based Prepaid Application Server • Handles multiple prepaid services simultaneously • Has more flexible credit control Messaging Services IMS-PSTN Call 2018/3/15 10

LABORATORY 117 Prepaid Messaging Delivery CCR: Credit Control Request CCA: Credit Control Answer 2018/3/15 SMS: Short Message Service 11

LABORATORY 117 IMS-PSTN Prepaid Call Setup CCR: Credit Control Request CCA: Credit Control Answer 2018/3/15 IAM: Initial Address Message ANM: Answer Message 12

LABORATORY 117 Prepaid Call Force-Termination CCR: Credit Control Request CCA: Credit Control Answer 2018/3/15 REL: Release Message RLC: Release Complete 13

LABORATORY 117 PAS Charging Policy Sending out a message during an IMS call may result in insufficient credit left for this ongoing call. A threshold XT is set to protect the in-progress IMS-to -PSTN call when admitting a prepaid message delivery: x*: the remaining credit left when a message arrives Tm: the charge for a prepaid instant message delivery x* -Tm≧ XT Send immediately x* -Tm< XT Delay until the call is finished 2018/3/15 14

LABORATORY 117 Analytic Output Measures Assuming 1 credit unit = 1 time unit (Tm) Message delivery delay is 0 Message delivery delay is t 3 -t 2 PUFT : the unnecessary force-termination probability of an inprogress call (i. e. , x>tc and x*-Tm < t 3 -t 2) E[td]: the expected delay for the message services 2018/3/15 15

LABORATORY 117 Input Parameters X: the amount of the initial prepaid credit lm: the arrival rate of the instant messages tc: the prepaid call holding time 1/m: the expected call holding time x: the remaining credit left when a call starts 2018/3/15 16

LABORATORY 117 Analytic Results [1/2] The pmf of the Poisson messaging arrivals during tc is Let fx(x) be the density function of the remaining credit x left when a prepaid call arrives. The UFT probability for XT =0 is derived as 2018/3/15 17

LABORATORY 117 Analytic Results [2/2] For XT ≥ X , the expected unnecessary delay for message delivery is derived as Mean Residual Life for tc is E[tc 2]/2 E[tc] 2018/3/15 18

LABORATORY 117 Effects of the Inter-message Arrival Time (X = 25 E[tc] and E[tc]=4 Tm) PUFT increases as lm increases. - more message deliveries are likely to occur during an in-progress call E[td] is insignificantly affected by lm. - messages are random observation points of prepaid call holding intervals 2018/3/15 19

LABORATORY 117 Effects of the Variance of Call Holding Time (X = 25 E[tc] and 1/λm = 0. 5 E[tc]) The performance of both PUFT and E[td] degrades as the variance Vc of the call holding time increases. 2018/3/15 20

LABORATORY 117 Effects of the Initial Prepaid Credit Amount (1/λm = 0. 5 E[tc]) Both PUFT and E[td] decrease as X increases. - more credit units left for a prepaid call - more likely that there is enough credit for both services 2018/3/15 21

LABORATORY 117 Summary We developed an IMS prepaid application server to handle prepaid calls and messaging services in UMTS/IMS. This application server can be used to accommodate existing Internet services in the mobile environment. A threshold XT is set to protect the in-progress prepaid call when admitting a prepaid message delivery. The XT value should be appropriately selected to adapt to various traffic patterns. 2018/3/15 22

LABORATORY 117 [2] Modeling Online Credit Reservation Procedure in OCS S. -I. Sou, H. -N. Hung, Y. -B. Lin, N. -F. Peng, and J. -Y. Jeng, Modeling Credit Reservation Procedure for UMTS Online Charging System. IEEE Transactions on Wireless Communications, 6(11): 4129 -4135, 2007. 2018/3/15 23

LABORATORY 117 Previous approach (in research topic [1] and before) - Real-time deduction in the service control point New Approach - Diameter Credit Reservation at the OCS 2018/3/15 24

LABORATORY 117 Diameter Credit Reservation Procedure qi qi 2018/3/15 For a type i session, each time the OCS grants qi credit units to the session. When these credit units are consumed, the OCS grants next qi credit units to the session. 25

LABORATORY 117 RTCR Mechanism We propose a Recharge Threshold-based Credit Reservation (RTCR) mechanism. In RTCR, when the balance of the user account at the OCS is below a recharge threshold Cmin, , the mechanism reminds the user to refill the prepaid account by sending a Recharge Warning, and rejects new session requests 2018/3/15 26

LABORATORY 117 RTCR Mechanism (cont. ) Cmin is set too small not enough credit is left user satisfaction is degraded Cmin is set too large new session requests are unnecessarily rejected the user is frequently asked to refill the account 2018/3/15 27

LABORATORY 117 Output Measures E[Ni]: the expected number of the reserve credit message exchanges executed during a type-i session. The larger the E[Ni] value, the higher the credit control message overhead. Pc: the completion probability that all in-progress sessions are finished. The larger the Pc value, the better the user satisfaction. E[Cd]: the expected unused credit units left after all existing sessions are finished. Cd ≥ 0 if all sessions are normally finished. Cd = 0, otherwise. 2018/3/15 28

LABORATORY 117 Input Parameters n: the number of types of session-based IMS services li: the inter-session arrival rate of type-i service th, i: the session holding time of type-i service 1/mi: the expected value of th, i qi: the amount of credit units that the OCS grants in each reserve credit message for a type-i session Cmin: the recharge threshold in RTCR 2018/3/15 29

LABORATORY 117 Derivation for E[Ni] The expected number of the reserve credit message exchanges executed during a type-i session When th, i is Exponential distributed, we have 2018/3/15 30

LABORATORY 117 Derivation for Pc and E[Cd] (n=1) [1/2] For n=1, at t 1, the total unused credit is where. , (1) Reserve (2) Remain (3) Send Recharge Warning 2018/3/15 31

LABORATORY 117 Derivation for Pc and E[Cd] (n=1) [2/2] The completion probability is derived as The unused credit is derived as 2018/3/15 32

LABORATORY 117 Derivation for Pc and E[Cd] (n=2) [1/2] Case I. One active type-1 session (with residual holding time tr, 1) Case II. One active type-2 session (with residual holding time is tr, 2) Case III. Two active sessions (with residual holding times tr, 1 and tr, 2) For sufficiently small qi, the completion probability Pc is derived as 2018/3/15 33

LABORATORY 117 Derivation for Pc and E[Cd] (n=2) [2/2] For sufficiently small qi, the expected unused credit E[Cd] is derived as 2018/3/15 34

LABORATORY 117 Effects of the Granted Credit Units E[Ni] is not affected by the threshold Cmin and the number of services types. E[Ni] decreases as qi increases. - The session holding time is Exponential distributed with parameter mi. - We observe qi ≥ 2. 5/mi, E[Ni] ≈1. 2018/3/15 35

LABORATORY 117 Effects of the Recharge Threshold (n = 2, l 1 = m 1 and l 2 = m 2 = 2 m 1 ) 1/m: the expected session holding time Pc increases as the threshold Cmin increases. E[Cd] increases as the threshold Cmin increases. - more unused credit units are available in the prepaid account. 2018/3/15 36

LABORATORY 117 Effects of the Number of Session Types (Cmin = 6/m, li = im 1 and mi = im 1) When the recharge warning is sent, the number of simultaneous in-progress service sessions increases as n increases. Two conflicting effects are observed: - more credit units will be consumed in these sessions - the “net” unused credit units that have granted to the sessions increase 2018/3/15 37

LABORATORY 117 Summary We proposed the threshold-based credit reservation mechanism. Analytic and simulation models are developed to investigate Pc and E[Cd]. Based on our study, the operation can choose appropriate parameters qi and Cmin for various traffic conditions. 2018/3/15 38

LABORATORY 117 [3] Reducing Credit Re-authorization Cost S. -I. Sou, Y. -B. Lin and J. -Y. Jeng, Reducing Credit Re-authorization Cost in UMTS Online Charging System. Accepted and to appear in IEEE Transactions on Wireless Communications. 2018/3/15 39

LABORATORY 117 Online Charging for GPRS Sessions The ABMF keeps the subscriber’s account data and controls the account balance. In a telecom network, the ABMF and the SBCF may physically reside at different (and possibly remote) locations. The message exchanges in the Rc interface may be expensive. 2018/3/15 40

LABORATORY 117 Credit Re-authorization (Basic Scheme) aiqu Debit ai(q-qu) ajqu 2018/3/15 Reserve aiq Reserve ajq Debit aj(q-qu) ai: the number of credit units charged for a time unit for class i session q: the time units granted in each credit reservation 41

LABORATORY 117 Credit Re-authorization (Threshold-based Scheme) aiqu ajq 2018/3/15 If aiqu≥ d (ajq), skips ABMF message and grants aiqu credit units. Otherwise, executes ABMF message and grants ajq credit units 42

LABORATORY 117 Input Parameters N: the number of the Qo. S classes that may change in a GPRS session ai: the number of credit units charged for every time unit in a class i session d: the threshold used by the SBCF to determine whether to interact with the ABMF or not q: the time units granted to the GPRS session in each credit reservation qu: the time units left at the end of the previous sub-session 1/ms: the expected holding time of a sub-session P 0: the probability that the GPRS session terminates at the end of a sub-session 2018/3/15 43

LABORATORY 117 Analytic Modeling: Timing Diagram After a random time, an in-progress session either terminates with probability P 0, or switches to another Qo. S class with probability (1 - P 0)/(N-1). Exponential random variable with mean 1/ms 2018/3/15 How much have I spend on this session? Consume C credit in this period 44

LABORATORY 117 Output Measures M: the expected number of ABMF message exchanges for a GPRS session. The smaller the M value, the lower the ABMF message overhead. C: the expected undebit credit units when a balance check occurs during an in-progress session. The smaller the C value, the more accurate the account balance reported by the OCS. We use (MB, CB) and (MT , CT) to represent the output measures of the Basic scheme and the Threshold-based scheme, respectively. 2018/3/15 45

LABORATORY 117 Output Measures for the Basic Scheme When q is Exponential distributed with l, the expected number of ABMF exchanges in a session for the basic scheme: Expected # of subsessions Expected # of ABMF messages in a subsession When a balance check occurs, the expected undebit credit units are: The time period from the last updated ABMF 2018/3/15 Expected credit units consumed in a time unit 46

LABORATORY 117 Effects of the Threshold Parameter (N=2, α 2=2α 1 and P 0=0. 01) q The basic scheme is not affected by δ. For the threshold-based scheme, MT increases and CT decreases as δ increases. - the performance of the threshold-based scheme is similar to that of the basic scheme when δ≥ 2. 5. 2018/3/15 47

LABORATORY 117 Summary We proposed a threshold-based scheme with parameter d to reduce the traffic signaling for the OCS credit re-authorization procedure. The threshold parameter d increases the number of ABMF message exchanges (MT ) increases the undebit credit units (CT) decreases Combining the results from our previous work [2], the mobile operator can select the appropriate d and q values for various traffic conditions. 2018/3/15 48

LABORATORY 117 Future Directions Performance of the Tariff Switch Mechanism During a service session, the tariff information may be changed when a specified event occurs (i. e. , a tariff switch is reached). It is important to set the tariff switch time appropriately such that the OCS can handle all the requests without delaying the service continuity. Policy and Charging Control Integration When the Qo. S policy control and the content-based charging functionalities are used as separate mechanisms, it will increase the interworking cost between the network nodes. Through the Policy and Charging Control (PCC), integration of Qo. S policy and charging rules can be realized in the IMS network or other bearer network (e. g. , WLAN, Wi. MAX). The design of the PCC architecture to support roaming among heterogeneous wireless network is for further study. 2018/3/15 49

LABORATORY 117 Q&A 2018/3/15 50