proof-of-concept-by-infosys

16
Proof of Concept: MATRIXX Online Charging & Policy Management Engine _____________________________________________________________________________________________ Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 1 of 16 Proof of Concept: MATRIXX Online Charging & Policy Management Engine Submitted to: MATRIXX Software Version No. 0.4 Authorized by Ian Williams © 2009 Infosys Technologies Limited. Strictly private and confidential.

Upload: brokergd

Post on 03-Apr-2015

122 views

Category:

Documents


4 download

TRANSCRIPT

Page 1: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 1 of 16

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

Submitted to:

MATRIXX Software

Version No. 0.4

Authorized by Ian Williams © 2009 Infosys Technologies Limited. Strictly private and confidential.

Page 2: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 2 of 16

CONTENTS

1. ABSTRACT ...................................................................................................................................................... 3

2. EXECUTIVE SUMMARY ................................................................................................................................... 4

3. MATRIXX SOLUTION OVERVIEW .................................................................................................................... 6

4. POC METHODOLOGY ...................................................................................................................................... 8

5. POC RESULTS OVERVIEW................................................................................................................................ 9

5.1 POC TEST RESULTS .............................................................................................................................................. 9

6. CONCLUSION ................................................................................................................................................ 13

7. APPENDIX..................................................................................................................................................... 14

7.1 TABLE 1: LIST OF SCENARIOS TESTED IN POC. .......................................................................................................... 14 7.2 TABLE 3: PRICE PLANS USED .............................................................................................................................. 15

7.2.1 Price Plan 1 – Simple SMS Rating Scenario .......................................................................................... 15 7.2.2 Price Plan 2 – Moderate GPRS Rating Scenario ................................................................................... 15 7.2.3 Complex Price Plan 3 – Complex Voice Rating Scenario ...................................................................... 15

7.3 HARDWARE USED ............................................................................................................................................. 16

Page 3: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 3 of 16

1. ABSTRACT

The business models for Telcos remain under unremitting pressure with the emergence of 21st Century Business Drivers that increase pressure on margin, delivery and customer service.

We see three key challenges that need software driven solutions in a cost effective way.

The explosion in the amount of mobile data traffic. Our customers are seeing year on year growth in excess of 100% in the transactions carried on their networks.

Bill-shock and policy management. Legislation changes and customer expectations mean that it’s no longer acceptable to catch up with billable transactions at some point in the future.

Telco 2.0 Business Models. Marketing departments must be able to monetize new services and compete with new entrants

It’s not realistic to just throw greater and greater amounts of hardware at these problems, smart solutions must be put in place. This paper describes our evaluation of the Matrixx Software and its ability to resolve these issues.

Page 4: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 4 of 16

2. EXECUTIVE SUMMARY

Today when we look for a Next-Generation Telecom billing solution, then the key attributes we are

looking are:

Compliance to future technologies – Support for all industry standard protocols like RADIUS,

DIAMETER, IPDR, HTTP for collection, which enables support for future technologies like

WiMAX, 3G, CDMA 1.x,LTE etc.

Real time support for pre-paid capabilities – Multiple session management, quota management

and balance management.

Post-paid – pre-paid convergence – Management of post-paid and pre-paid accounts under the

same customer with balance transfer support.

With the number of quality telecom service providers increasing globally, acquiring new customers and

retaining the existing base is the biggest challenge for service providers. So the understanding of

customer usage patterns is the key for communication service providers in identifying the customer pain

points and possible areas of enhancements.

Statistics indicate that CSP’s may be losing an estimated 3% -11% of their revenue due to operation

leakages from network failure to create records, corrupt Call Detail Record data, delays in processing,

fraud, missing files, rating inaccuracy, collection problems, billing errors, prepay faults, interconnect

problems, software updates, provisioning errors and Debt/write-off.

Also with the advent of 3GPP, billing systems need to overcome the above problems as soon as possible.

MATRIXX OC/PM engine is one such solution which is 3GPP compliant and performs online charging,

account balance management, and rating and complies with the Diameter standard for authorization,

authentication, and accounting. This whitepaper is an attempt in analysing the proof of concept carried

out for this solution and provides the thorough analysis on the same. Various attributes for which the

Engine has been tested under the POC include CPU utilization, latency, throughput and performance

measured in terms of transactions per second.

Page 5: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 5 of 16

This whitepaper documents our evaluation of the MATRIXX OC/PM Engine against these key

performance indicators and also measures it’s capability of supporting high volumes of prepaid and

postpaid usage.

We have carried out our evaluation across a number of business scenarios, testing the scalability of the

node based architecture of MATRIX OC/PM engine and it’s response to high volume of events for

various subscriber bases with increasing levels of pricing complexity.

The results we achieved are very exciting. We have been able to demonstrate real-time rating in excess

of 10,000 transactions per second, per blade, with linear scalability in terms of both rating complexity

and hardware utilisation.

Page 6: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 6 of 16

3. MATRIXX SOLUTION OVERVIEW

MATRIXX OC/PM Engine is a 3GPP compliant modern online charging and policy management

system designed to support high volumes of prepaid and postpaid usage. Built on our patent-

pending Parallel-MATRIXX™ Technology, it combines extremely efficient transaction processing

with a highly flexible pricing, rating, and policy engine.

Key Functionality: MATRIXX OC/PM Engine encompasses the following functionality.

1. Real-time Rating and Charging: MATRIXX OC/PM Engine supports real-time

authorizations, re-authorizations, and session management, and can rate events both

online and offline. The prices defined in the pricing catalog are mapped to a multi-

dimensional array of algebraic formulas that are implemented at the system level and

isolated from the business logic, which makes processing extremely fast. The algebraic

equations enable MATRIXX OC/PM Engine to rate complex pricing structures and still

achieve ultra-high performance.

2. Pricing: The MATRIXX OC/PM Engine can handle complex charging and discounting

structures so service providers are not limited to basic pricing plans. IT personnel can

easily introduce new rating sequences, as frequently as needed, without requiring

updates to the core system to incorporate the changes. MATRIXX OC/PM Engine

provides an intuitive graphical user interface that allows you to easily create elaborate

charging and discounting models and reuse them across products and services. There is

no tradeoff between pricing complexity and performance, so more data can be

processed without reducing performance or decreasing efficiency. Pricing

administrators can use the templates to set up elaborate pricing structures and save the

configurations for reuse across product catalogs. This provides a set up once, reuse

anywhere approach that makes the MATRIXX Catalog Builder unique to other pricing

applications.

3. Subscriber and Balance Management: MATRIXX OC/PM Engine provides a sophisticated

set of balance management features so customers can share or allocate balances among

devices and subscribers. Integrated balance reservations ensure risk-free balance

sharing with zero exposure to revenue loss or leakage. You can set credit limits and

Page 7: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 7 of 16

other thresholds on balances and trigger notifications to occur when a threshold is

crossed. For example, you can trigger a threshold notification to warn subscribers about

impending charges.

4. Policy Management: MATRIXX subscriber policy management enforces business rules

set up by service providers or subscribers to control access to services and balances.

System Highlights: There are several system highlights that set MATRIXX OC/PM Engine apart

from other online rating and charging systems.

1. Transaction Processing: The Parallel-MATRIXX transaction control architecture removes

the overhead involved in tracking data throughout the commit process (including

tracking any other processes that want to access that data). This allows it to rate

thousands of events concurrently while guaranteeing data integrity. The Parallel-

MATRIXX Clustering architecture allows identical copies of data to be distributed across

the MATRIXX OC/PM Engine and to be owned equally by each OC/PM blade in the blade

enclosure. Shared ownership of all data removes the chance of a single point of failure,

which is common in most distributed database management systems. MATRIXX OC/PM

Engine is comprised of several identical OC/PM blades. Each OC/PM blade is fully

contained on one blade server and is able to process events at full speed. The OC/PM

blade redundancy creates a highly available system that can handle an incredibly large

throughput without compromising performance. Adding more blades to MATRIXX

OC/PM Engine further increases the processing power.

2. High Availability: MATRIXX OC/PM Engine is comprised of several blade servers that are

identical in architecture and can process events independently. Each blade server

contains the same data set, so if one blade server goes offline, the other blade servers

take over processing for it. This guarantees high availability of MATRIXX OC/PM Engine.

3. Simple Configuration: You do not need to write complex code to configure MATRIXX

OC/PM Engine behavior and functionality. Instead, to configure MATRIXX OC/PM

Engine, you use XML specifications and a graphical user interface. This makes it

extremely easy to change the current configuration, such as configuring system-wide

parameters, network-to-MATRIXX data mapping, balance types, and pricing

components.

Page 8: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 8 of 16

4. POC METHODOLOGY

Our Proof of Concept for MATRIXX Online Charging & Policy Management Engine attempts to validate

the behaviour of MATRIXX OC/PM engine for the following parameters:

CPU Utilisation: This is defined as the percentage of available CPU processing cycles that are used

for any reason during the benchmark run. We were looking to ensure that the benchmark ran at

operational CPU loads (< 60%) to validate how realistic the results are.

Latency: We measured the elapsed time from receipt of the diameter request message at the

MATRIXX diameter gateway and the transmission of the diameter response message by the

MATRIXX diameter gateway. This covers the entire processing of the event, including the charge

calculation, transactional balance updates, and full synchronous logging.

Performance (Transactions/sec): We recorded the number of diameter charge request messages

that were fully processed and responded to per second averaged over the entire benchmark run.

Each test was carried out in real-time with synchronous logging of events and with full, ACID-compliant

transactions. We created a number of business scenarios whereby we used a two dimensional based

approach involving increasing pricing complexity and increasing numbers of subscribers. The scenarios

had different numbers of subscribers and rate plans of 1, 5 and 10 dimensions. Each scenario was run

with 1, 2 and 4 blades.

For each test, we pre-loaded all of the subscriber information, as we were not seeking to evaluate this

part of the system. Using a SEAGULL diameter call simulator, we prepared random samples of data for

each of the rating dimensions being tested and produced files containing the appropriate number of

events for each test.

Once the file had been prepared, we started the charging mechanism and logging and used the log files

to populate our test results.

Page 9: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 9 of 16

5. POC RESULTS OVERVIEW

5.1 POC TEST RESULTS

5.1.1 LINEAR SCALABILITY OF MATRIXX OC/PM ENGINE

Figure 1 shows the graph depicting Linear Scalability of Matrixx OC/PM engine whereby for a

particular price plan (Complex Voice Plan), Performance and CPU utilization are measured by varying

the number of blades for various subscriber bases.

Key observations noted here are:

Performance (events per second) is directly proportional to the number of blades used.

Max CPU Utilization, which can also be taken as a measure of Peak Load goes up if we

increase number of blades keeping other parameters constant.

Max CPU utilization is not increasing steeply and remains within our 60% threshold.

Figure 1. Linear Scalability

Page 10: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 10 of 16

5.1.2 LATENCY VS. THROUGHPUT KEEPING OTHER PARAMETERS CONSTANT

Figure 2 shows the graph between Latency and Throughput variations keeping other parameters

constant. Following are the key observations during our POC:

Latency decreases with increasing number of blades.

Also follows the Industry standard whereby 98% of calls have latency of around 15 ms. So,

this shows that software is in compliance with existing standards.

Figure 2 Graph showing Latency vs Throughput

Page 11: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 11 of 16

5.1.3 LATENCY VS. SUBSCRIBER BASE KEEPING OTHER PARAMETERS CONSTANT

Figure 3 shows the graph between Latency and Subscriber Base keeping other parameters

constant. Following are the key observations during our POC:

Latency is proportional to subscriber base if number of blades is constant.

Figure 3 Graph showing Latency vs. Subscriber base

Page 12: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 12 of 16

5.1.4 LATENCY VS. PRICING COMPLEXITY KEEPING OTHER PARAMETERS CONSTANT

Figure 4 shows the graph between Latency and Pricing complexity keeping other parameters

constant. Following are the key observations during our POC:

There is very little change in the latency when there is increased complexity with the price

plans.

Figure 4 Graph showing Latency vs. Pricing Complexity

Page 13: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 13 of 16

6. CONCLUSION

Our results were very much in line with our expectations, matching Matrixx’s predictions. The headline result of 10,000 transactions per second, per blade, was achieved with all combinations of price plans and numbers of subscribers. This demonstration of linear scalability combined with high performance gives a new solution to the 21st Century challenges we and our customers have identified. Also a key driver of this Engine is about maintaining latency at high loads as well.

Page 14: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 14 of 16

7. APPENDIX

7.1 TABLE 1: SCENARIOS AND RESULTS IN POC.

47.718 47.751 47.753

10,478 10,471 10,470

19.15% 22.92% 26.84%

19.52% 23.37% 28.82%

23.912 23.913 23.913

10,455 10,454 20,909

33.01% 30.81% 31.91%

36.05% 32.34% 34.20%

23.914 23.915 23.915

10,454 10,453 20,907

31.65% 30.29% 30.97%

32.33% 31.87% 32.10%

23.914 23.913 23.914

10,454 10,454 20,908

35.09% 28.54% 31.82%

37.45% 30.10% 33.78%

12.008 12.009 12.009 12.009

10,409 10,408 10,408 10,408

40.63% 37.90% 42.95% 38.77%

43.16% 40.32% 45.16% 39.20%

11.974 12.009 12.011 12.012

10,439 10,408 10,407 10,406

32.58% 38.73% 39.41% 48.11%

42.92% 49.51% 43.14% 49.31%

12.002 12.002 12.002 12.002

10,415 10,415 10,415 10,415

38.08% 34.38% 35.68% 37.79%

38.42% 35.48% 41.22% 40.24%

476.322

10,497

22.54%

23.35%

238.309 238.308 238.309

10,490 10,490 20,980

Max CPU Util. = Max CPU Util. = Max CPU Util. =

Max CPU Util. =

Max CPU Util. = Max CPU Util. = Max CPU Util. = Max CPU Util. =

Max CPU Util. = Max CPU Util. =

Max CPU Util. = Max CPU Util. = Max CPU Util. = Max CPU Util. = Max CPU Util. =

Max CPU Util. =

Max CPU Util. = Max CPU Util. = Max CPU Util. =

Max CPU Util. = Max CPU Util. = Max CPU Util. =

Max CPU Util. = Max CPU Util. = Max CPU Util. =

Average CPU Util. =

Average CPU Util. =

Average CPU Util. = Average CPU Util. =

Average CPU Util. = Average CPU Util. = Average CPU Util. = Average CPU Util. =

Average CPU Util. = Average CPU Util. = Average CPU Util. =

Average CPU Util. = Average CPU Util. = Average CPU Util. =

Average CPU Util. =

Average CPU Util. = Average CPU Util. = Average CPU Util. =

Average CPU Util. = Average CPU Util. =

Average CPU Util. = Average CPU Util. = Average CPU Util. =

Average CPU Util. = Average CPU Util. =

Events / Sec = Events / Sec = Events / Sec =

Events / Sec = Events / Sec = Events / Sec = Events / Sec = Events / Sec =

Events / Sec =

Events / Sec =

Events / Sec = Events / Sec = Events / Sec = Events / Sec = Events / Sec =

Events / Sec = Events / Sec =

Events / Sec = Events / Sec = Events / Sec = Events / Sec =

Events / Sec = Events / Sec = Events / Sec =

Events / Sec = Events / Sec = Events / Sec =

Events / Sec =

Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

C-Total

C-1

C-1 C-2 C-Total

M-4 M-Total

C-1 C-2 C-3 C-4

E-3 E-4 E-Total

M-1 M-2 M-3

M-Total

C-1 C-2 C-Total

E-1 E-2

2 Blades

E-1 E-2 E-Total

M-1 M-2

2 Blades

2 Blades

4 Blades

4 Blades

4 Blades

1 Blade

1M Subs / 5M Events / FSL

1M Subs / 5M Events / Full Synchronous Logging

100K Subs / 500K Events / Full Synchronous Logging

100K Subs / 500K Events / Full Synchronous Logging

100K Subs / 500K Events / Full Synchronous Logging

100K Subs / 500K Events / Full Synchronous Logging

100K Subs / 500K Events / Full Synchronous Logging

100K Subs / 500K Events / Full Synchronous Logging

2 Blades

Max CPU Util. = Max CPU Util. = Max CPU Util. =

Events / Sec = Events / Sec = Events / Sec =

Average CPU Util. = Average CPU Util. = Average CPU Util. =

100K Subs / 500K Events / Full Synchronous Logging

1 Blade

E-1 M-1 C-1

Elapsed time (sec) Elapsed time (sec) Elapsed time (sec)

Page 15: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 15 of 16

7.2 PRICE PLANS USED

7.2.1 PRICE PLAN 1 – SIMPLE SMS RATING SCENARIO

This price plan charges subscribers based on the number of SMS messages consumed over the past month. The rate

is a simple per-message flat fee.

7.2.2 PRICE PLAN 2 – MODERATE GPRS RATING SCENARIO

This price plan charges subscribers for GPRS service. Subscribers prepay for a megabytes allowance and a roaming

megabytes allowance to which charges are applied. If they go over their allotment of megabytes, overage charges

apply.

The rates are based on the following 5 rating parameters.

Roaming or not roaming. If roaming, charges are based on the country in which the usage occurs (Zone A,

B, C, D).

Content type – email, text message, or general Internet usage.

Device type – blackberry device or other smart phone.

Prepaid data balance – if gone, overage charges apply.

Prepaid roaming data balance – if gone, overage charges apply.

7.2.3 COMPLEX PRICE PLAN 3 – COMPLEX VOICE RATING SCENARIO

This price plan charges subscribers for Voice service based on the number of minutes consumed over the past

month.

The rates charge a different amount based on the following 10 rating parameter and the values that are valid at the

time of rating.

Time-of-day – peak period, off-peak period, or weekend calling.

Calling zone – local, long distance, or international calling to Asia, Latin America, or Europe.

Monthly Usage Balance – if the balance is over the C$500 threshold, rates change.

Carrier ID – on network or off-network.

Friends and family – in calling circle or out of calling circle.

Discounted minutes balance – if available, subtract from this balance and charge a different rate.

Subscriber’s birthday – if it is a birthday, the subscriber is charged a different rate.

Holiday Rates – if it is a holiday, the subscriber is charged a different rate.

Roaming or not roaming rates.

The SMS usage total – if the subscriber has sent over 50 SMS, the call is charged a different rate.

Page 16: proof-of-concept-by-infosys

Proof of Concept: MATRIXX Online Charging & Policy Management Engine

_____________________________________________________________________________________________

Whitepaper by Infosys on POC: MATRIXX Online Charging & Policy Management Engine Page 16 of 16

7.3 HARDWARE USED

HP

Qty Product Description

1 507019-B21 HP BLc7000 CTO 3 IN LCD ROHS Encl

4 507778-B21 HP BL460c G6 X5550 1P Svr

4 507793-B21 HP X5550 BL460c G6 FIO Kit

48 500658-B21 HP 4GB 2Rx4 PC3-10600R-9 Kit

8 507750-B21 HP 500GB 3G SATA 7.2K 2.5in MDL HDD

1 455880-B21 HP BLc VC Flex-10 Enet Module Opt

2 AT004A HP P4500 1.8TB SAS Storage System

1 J9145A HP ProCurve 2910al-24G Switch

Plot No. 44 & 97A, Electronics City , Hosur Road, Bangalore - 560 100 Phone: +91 80 28520261 Fax: +91 80 28520362