You are on page 1of 10

Openness in AXE

Einar Wennmyr

In recent years, the term open system has become a catch phrase in the • network element management (NEM)
data and telecommunications industries. Open systems are expected to protocols such as Telnet, file transfer pro-
improve time to market and make it easier for operators to introduce new tocol (FTP), file transfer, access and man-
features that will help them to attract and retain their subscribers. As agement (FTAM), and so on;
competition in the market stiffens, it matters more and more who is first to • asynchronous transfer mode (ATM) pro-
tocols; and
market with new applications.
• Internet protocols (IP).
In this article, the author describes various attributes of open systems, Obviously, AXE is also continually being
and to what extent Ericsson has introduced them into AXE. He also updated to accommodate new standards.
describes the added value that is being integrated into AXE, extending its Some examples include the implementation
competitive edge even further. of the bearer independent call control
(BICC) protocol, the media gateway control
protocol (H.248), and the common object
What is openness? request broker architecture (CORBA),
which can be used for implementing
A distinction can be made between external Ericsson’s integration reference points (IRP)
openness to a node (network openness) and for operation and maintenance (O&M).
openness within a node (system openness).
System openness
Network openness System openness signifies the use of stan-
Network openness signifies the ability to in- dard, commercially available components to
teroperate with other nodes in different net- build the AXE system platform. More
works. In this context, the AXE system has specifically, system openness is attained
always been open—it supports numerous through the use of
protocol standards and market variants, and • commercial hardware components (sub-
can interoperate with any node implement- racks, boards, chipsets);
ed on another system platform, provided that • standard hardware building practices and
node supports the same protocols. Examples buses;
of protocols currently supported by AXE are: • standard programming languages using
• channel-associated signaling (CAS); standard software-development tools—
• common channel signaling (CCS) and ap- software can thus be ported to different
plications on top of it, such as ISDN sig- hardware and operating systems; and
naling user part (ISUP), the mobile ap- • commercially available software compo-
plication protocol (MAP), and so on; nents and interfaces.
In the past, AXE was considered a propri-
etary system with few standard components.
Today, however, this description of AXE no
Figure 1
System unbundling in AXE. longer applies, since standard components
are being introduced at an increasingly ac-
celerated pace. By using standard compo-
Sales and nents to build their systems, companies like
distribution Ericsson can concentrate on their core busi-
ness and still benefit from technological ad-
vances in other segments. What is more,
Application
software they can achieve shorter time to market
Switching Operation and Routing (TTM). The sourcing of components should
maintenance
be managed with care, however. And in
some cases, sourcing might not be the ap-
System propriate or viable alternative:
software APZ 21 APM NT OSE-Delta
• Satisfactory commercial components
might not be available on the market.
• It might be more profitable to develop
Hardware CPG APG 40 IPNA GS DEV RPP and produce components in-house—even
when similar components are available on
the market.
• Proprietary, in-house designs might be
Chips Custom Pentium Alpha needed to remain competitive.
• Some “open” components lack the inter-
faces that are needed for integration.

32 Ericsson Review No. 1, 2000


Components in AXE ET 155 STM ET CE

Figure 1 gives a component view of AXE.


The goal is to unbundle the system into a IOG RP RP RPP RPP RPP ATM ET 622
layered set of cooperating (interoperating)
components—the trend we are seeing in the RPB-S
market is for specialization in horizontal
product layers. By pursuing this approach,
CPx CPz
Ericsson can PLEX
CPy ATM
• take full advantage of commercially avail-
able technology; IPN
• reuse existing components and reduce the
time it takes to introduce new products;
and AP AP Hard disk/
• separate software functions from the hard- optical disk
ware implementation.
Figure 2 gives an overview of the evolving GUI
architecture of the AXE system. Several of
the products shown are described in this
article. Figure 2
AXE system architecture evolution.
The adjunct processor
platform processors with the following type of func-
The adjunct processor (AP) is one of the first tionality:
applications in AXE to use large building • input-output (I/O) functions (a file sys-
blocks from a commercial vendor.1 The AP tem, storage, formatting and output of
complements the central and regional call data records for charging, and storage

BOX A, ABBREVIATIONS

AM Application module GDDM-H Generic device and datacom OMG Object Management Group
AOT Ahead of time magazine, half-height ORB Object request broker
AP Adjunct processor GPRS General packet radio service OSS Operations support system
APIO Adjunct processor input and GSM Global system for mobile PAL Privileged architecture library
output communication PCI Peripheral component
APSI Application platform sevice HDLC High-level data link control interconnect
interface IDL Interface description language PCU Packet control unit
ASA Assembly statements IIOP Internet inter-ORB protocol PMC PCI mezzanine card
ASIC Application-specific integrated circuit IN Intelligent network RMP Resource module platform
ATM Asynchronous transfer mode I/O Input-output RNC Radio network control
BICC Bearer independent call control IP Internet protocol RNS Radio network server
BMC Base management controller IPC Interprocessor communication RP Regional processor
BSC Base station controller IPN Interplatform network RPC Remote procedure call
CAS Channel associated signaling IPNA IPN adapter RPH RP handler
CCS Common channel signaling IPU Instruction processing unit RPHM RPH magazine
CORBA Common object request broker IRP Integration reference point RPHMI RPHM interface
architecture ISDN Integrated services digital RPP Regional processor platform
CP Central processor network SMP Symmetric multiprocessor
CSH Connection service handler ISUP ISDN signaling user part SPU Signal processing unit
DAT Digital audio tape JIT Just in time STOC Signaling terminal open
DDS Digital data storage LAN Local area network communication
DMA Direct memory access MAP Mobile application part STS Statistics service
DSP Digital signal processor MAU Maintenance unit TCP Transmission control protocol
ENGINE Next-generation switch MML Man-machine language TDM Time-division multiplexing
EPSB Ethernet packet switch board MSCS Microsoft cluster server TRH Transaction record handler
ET Exchange terminal MTBF Mean time between failures TTM Time to market
ETC Exchange terminal ciruit NEM Network element management UDP User datagram protocol
ETCE Exchange terminal circuit emulation NSP Next-generation switch platform UMTS Universal mobile
FOS Formatting and output service NT Network termination telecommunications system
FTP File transfer protocol O&M Operation and maintenance UPB Update board
FTAM File transfer, access and OCITS Open communication Internet VM Virtual machine
management transport service XSS Existing source system

Ericsson Review No. 1, 2000 33


of counters for statistics and traffic mea- and stored persistently. Events are also
APIO FOS STS ... surements); forwarded for alarm processing;
• boot server for the AXE central processor; • alarm handling;
• support for man-machine communica- • support for upgrading software in the
tion; and AP—high-level commands enable oper-
CP <––> AP Process supervision, • connectivity to operations support sys- ators to initiate a software upgrade, com-
communication event and alarm
handling, ... tems (OSS) including protocols for file plete it, and if necessary, to fall back on

transfer, message transfer and operator ac- old software;
cess. • authorization; and
Microsoft Standard The next-generation computer platform for • audit log—all commands issued to AXE
Cluster communication
Server protocols the adjunct processor functions is called and all man-machine-language (MML)
APG 40. Like its predecessor (the APG 30), printouts are logged, as are all state
the APG 40 will be based on open, com- changes in the network termination (NT)
Windows NT operating system mercial products. part.

The CP-AP communication service, statis-
Focus of development tics service (STS), formatting and output
Pentium- and cPCI-based hardware platform Ericsson has focused its development efforts service (FOS) for charging data, and the AP

on the interface to the central processor and input-output (APIO) for backup are each
on improving operator support for handling proprietary solutions.
Figure 3 the adjunct processor (Figure 3). In partic-
Component view of the adjunct processor ular, developers have worked on improving Sourced products
(AP).
or adding the following functionality: To a large extent, the platform requirements
• CP-AP heartbeat; are fulfilled by the following commercial
• system calendar synchronization between products:
the AP and CP; • A commercial processor platform with
• system monitoring and diagnostics— sufficient capacity and I/O devices. This
hardware events, system log analysis, Mi- large building block, which has been
crosoft Cluster Server (MSCS) process su- sourced from an external vendor, is based
pervision; on Pentium II processors. It contains one
• event handling—every reported event is or three 18 Gbyte hard disks (depending
supplemented with a date and time stamp on the configuration), digital audio tape
(DAT) or digital data storage (DDS)
streaming tape drivers, 100 Mbit/s Eth-
ernet ports, PCI mezzanine card (PMC)
modules for communication support, and
Figure 4 power supply.
AXE architecture with the regional processor with PCI bus (RPP). • Microsoft Windows NT operating sys-
tem. Many third-party software vendors
High-speed WAN High-speed LAN
E1/T1 WAN provide products for this operating sys-
tem.
PMC
• Microsoft Cluster Server (MSCS) software
ETC for high-availability servers. The MSCS
RPP
currently supports a two-node cluster,
defining the interdependency between,
RPP and handling the restarting of, services,
EPSB thus making the entire system fail-safe.
RPP • Disk management and data handling—
Volume Manager supports online man-
ETC agement, reconfiguration, and fast-failure
Group switch
Access recovery; Diskkeeper handles defragmen-
ETC
network RPP tation; WinZip handles the compression
ETC of data; backup software handles backup
RPP EPSB and disk cloning.
• Connectivity to OSS—several commer-
cial protocols are used, including the re-
RPB-S
mote procedure call (RPC), the transmis-
CP
sion control protocol/Internet protocol
(TCP/IP), and the Internet inter-ORB
protocol (IIOP).

34 Ericsson Review No. 1, 2000


Data communication -48V
interfaces in AXE PMCÞcarrier board
Optional
Other open products already included in the
AXE system are the regional processor with
PCI bus (RPP) and the Ethernet packet PCI
switch board (EPSB). The RPP, which sup- -48V
ports data communication-related telecom- I/O board RPB-S
Mandatory M-bus
munication applications, offers a range of 2 x 100 Base-Tx
open hardware interfaces, software applica- DL2
tions, and a complete development envi- PCI 2Mbit/s TDM Backplane
ronment. One of the first applications to use interfaces
the RPP and EPSB is the packet control unit -48V
CPUÞboard
(PCU) in the base station controller (BSC). Mandatory

The PCU provides support for general pack- DL2
et radio service (GPRS) in GSM networks.2
PCI 2Mbit/s TDM
RPP -48V

The RPP, which is situated as an ordinary DSPÞboard


Optional
regional processor (RP) in the AXE archi-
DL2
tecture (Figure 4), extends the functionali-
ty of a traditional regional processor by sup-
porting several protocols and physical links
and by providing support via duplicated Figure 5
Structure of the regional processor with PCI bus (RPP).
Ethernet for distributing a datacom appli-
cation over multiple RPPs. Both sourced
and proprietary components have been used
in the RPP (Figure 5).
The RPP CPU is based on commercially
available processors (currently a 333 MHz
PowerPC). Greater capacity is achieved by The different components within the RPP
upgrading the RPP CPU as new processor are connected internally via a PCI bus.
generations are introduced. The basic
processor operating system is OSE Delta. Ethernet packet switch board
Some additions have been introduced on The Ethernet packet switch board and Eth-
top of the OS, yielding a product called ernet connectors in the backplane of the
RTOS (real-time OS). Standard program- housing subrack (GDDM-H) have been in-
ming languages, such as C and C++, can be troduced to support applications that are
used to build datacom applications on top distributed over multiple RPPs. Ethernet
of RTOS. This makes it easy to source ap- was chosen because it is the industry stan-
plications and to find development re- dard for local area network (LAN) commu-
sources in the future. nication.
The I/O board contains standard Ethernet The Ethernet switch, which is built on a
interfaces and custom interfaces to AXE (to single board using sourced switch circuits,
the group switch and the RP bus). Standard gives interworking functionality to separate
networking interfaces can also be included RPPs in the same subrack, in different sub-
on externally sourced PMC modules on the racks, or to external equipment. The Ether-
PMC carrier board. Various commercially net switch contains thirteen 10 Mbit/s ports
available digital signal processors (DSP) can and three 100 Mbit/s ports. The 10 Mbit/s
be included for bit-stream-oriented proto- ports and one of the 100 Mbit/s ports are
cols. Most protocol support for modem, fax, available in the backplane. The remaining
V.110, voice coding, echo cancellation, two 100 Mbit/s ports are available at the
ATM, IP (TCP/IP) and high-level data link front of the board. The switch is duplicated
control (HDLC) are provided through for redundancy.
sourced products.
O&M capabilities are primarily based on
extensions of traditional AXE functionality
Interplatform network
in the central processor. However, some The interplatform network (IPN) is a prod-
maintenance can be performed via TCP/IP. uct that will introduce a 100 Mbit/s Ether-

Ericsson Review No. 1, 2000 35


CP-1 CP-2 AP-1 AP-2 Other system The IPN will first be deployed in AXE-
A B A B A B A B 1 2 3 based nodes of the universal mobile telecom-
munications system (UMTS).

Benefits of the IPN


IPN functionally replaces the signaling ter-
minal for open communication (STOC),
which currently terminates Ethernet in
IPN AXE. Note: the STOC solely terminates
side A
10 Mbit/s Ethernet. Other benefits of the
IPN
IPN are increased bandwidth for data trans-
side B fer out of the CP, improved performance for
sending large volumes of data to and from
the CP, and improved latency and through-
put of messages.
Increased bandwidth
Figure 6
The interplatform network (IPN). Greater bandwidth for transferring data out
of the CP is obtained by connecting the IPN
to the RPH bus instead of to the RP bus.
net interface into the AXE CP. An impor- The IPN can potentially use the entire RPH
tant reason for choosing the 100 Mbit/s Eth- bus bandwidth of 160 Mbit/s instead of
ernet interface is that it is an industry stan- being limited to the RP bus bandwidth of
dard supported by many commercial prod- 10 Mbit/s—notwithstanding, this band-
ucts, and as such, simplifies interconnecting width must be shared with the RP buses
AXE with other commercial products. As needed for ordinary RPs. The distribution
shown in Figure 6, the IPN has many ap- of bandwidth can be configured and set by
plications: operator commands. In the next-generation
• CP-AP communication—improved band- processor, Ethernet will be terminated di-
width can decrease the time required for rectly on a CP board instead of via the RPH
backup and reloading software; bus.
• communication between the AXE CP and
other platforms (such as the next- Improved performance
generation switch platform, NSP, or Support for soft, direct-memory access
AXD 301) for hybrid nodes or for mov- (DMA) in the CP microprogram improves
ing functionality from AXE, in order to performance when large volumes of data are
gain an increase in capacity3, 4 and sent to and from the central processor (64
• CP-CP communication within a clus- Kbytes of data can be transferred at one
ter—this alternative is currently being in- time). This makes for the efficient transfer
vestigated as an option for providing of communication buffers and dynamic
greater capacity in large nodes. buffers in AXE to the IPN adapter (IPNA).

Figure 7
Protocol stacks in the IPNA. CP IPNA AP

OCITS OCITS
Gateway TCP/IP
Gateway TCP/IP
RPHB Ethernet
RPHB Ethernet

RPHB IPN

36 Ericsson Review No. 1, 2000


RPB-P RPB-P
Ethernet Ethernet
RPB-S RPB-S
AP AP
RPHP RPHP
IPNA IPNA
RPHS RPHS Ethernet
RPIO RPIO Ethernet
RPHM-A RPHM-B

RPHB and TAB RPHB and TAB

CP state info
Data transfer

CPU-A CPU-B

RPHB SPU IPU UPB


interface board

System bus
Compact PCI Compact PCI

BMC and
MUX/ Memory
module Power
DEMUX
Figure 8
The components and data flow in the APZ
212 40 CP.

Improved latency and throughput APZ 212 40


Latency and throughput of messages are es- Ericsson’s APZ 212 40, which is the next-
pecially important when the IPN is used for generation central processor in AXE, will be
traffic applications. These two characteris- based on a commercial processor and not on
tics will be improved using a commercial, in-house processor chips. This decision was
high-capacity processor chip that can be up- based on a benchmark test of several com-
graded as new generations become available. mercial processors. A compiler prototype
has been developed for the most promising
IPNA design candidate, in order to test the execution of
The IPN connects to the AXE CP through various telecommunications applications.
the IPN adapter, which will be based on The results compare favorably to those of a
• a standard, commercial processor with the traditional CP and indicate that the com-
OSE Delta operating system; and mercial processor delivers roughly the same
• a commercial Ethernet chip. capacity as can be derived from a next-
Proprietary ASICs will be developed to sup- generation, in-house solution.
port the physical interface to the RPH bus
(Figure 7). Interfacing the processor to the AXE
The IPNA terminates the TCP/IP proto- architecture
col and includes gateway functionality for Ericsson will develop several hardware and
converting TCP/IP into internal AXE pro- software components (Figure 8)
tocols. The file transfer protocol will be used • to guarantee fault tolerance (hardware and
for transferring data to the AP. Other stan- software);
dard protocols can be provided if necessary. • to provide support for upgrading software
The open communication Internet transport while traffic is being handled in the
service (OCITS) layer provides additional processor; and
support for the application level protocol, • to provide support for existing interfaces
such as addressing, message fragmentation to other parts of the AXE system.
and recombination. The CPU subrack will contain two CPU

Ericsson Review No. 1, 2000 37


IPU SPU

APZ CP OS AOT compiler

JIT compiler APZ VM

Commercial OS

Alpha PAL code Alpha PAL code


Interconnect to
other CP side
Alpha silicon Alpha silicon

Memory

Execution process, IPU thread


Execution process, SPU thread
Figure 9 Compilation process (JIT shares thread with APZ VM)
CP layering.

boards (the instruction processing unit, stead, a cold-standby processor will be used.
IPU, and the signal processing unit, SPU) This means that during ordinary traffic, the
and up to 16 Gbytes of memory. It will also standby CPU will be loaded with the latest
contain maintenance unit (MAU) function- version of the software and configuration
ality and an interface board to the RP bus data, but will not be updated with traffic
handlers for interfacing existing RPs and data. If a non-recoverable hardware failure
APT hardware. occurs, the standby processor will take over
Fault tolerance will be supported through execution by reloading and restarting. Be-
a duplicated CPU. However, in contrast to cause the mean time between failures
previous generations, the execution side and (MTBF) is very long for modern processors,
the standby side of the APZ 212 40 will not the MTBF goals for AXE can still be met.
operate in parallel-synchronous mode; in- The update board (UPB) will implement
the physical link between the two CP sides
through 1 Gbit/s Ethernet. In addition,
100 Mbit/s Ethernet will be supported for
AP communication (backup, reload, charg-
ing data, and so on).
Figure 10
System modules.
Equipment practice
The APZ 212 40 will fit into the standard
BYB 501 equipment practice. Two CP sides
XSS AM AM AM fit into a standard-width subrack; each CP
side will be built using the standard, com-
pact PCI 6U equipment practice. Mainte-
nance buses will be based on the I2C stan-
dard and kept independent of the cPCI bus.

RMP CP structure
The memory chip, processor chip, privi-
leged architecture library (PAL) code, and
operating system (Compaq Tru64 UNIX, a
APZ commercial 64-bit version of UNIX) will be
sourced from external vendors.
Ericsson will provide the components and

38 Ericsson Review No. 1, 2000


functionality indicated in the shaded areas ity. Therefore, more capacity will become
of Figure 9. The APZ virtual machine (VM), available in AXE if the AMs can be exe-
which is implemented in C++, provides the cuted on separate platforms. Some exam-
PLEX execution environment as well as ples where this approach has been pursued
middleware support for function changes are: moving IN functionality and ISUP
and switching between CP sides (in case of protocol termination to separate AXE for
failure). The APZ VM executes the PLEX ENGINE, moving the VLR functionali-
software in a process with two threads: the ty from AXE to NSP for TDMA, and fi-
IPU thread and the SPU thread. The process nally investigating whether IN function-
provides a scheduler, job buffer handling, ality can be ported to the NSP in a
communication support (for sending and re- prototype.
ceiving signals), and support for error han- • step-by-step migration—if an applica-
dling, recovery, and a run-time log. tion module (or part of one) needs to be
The traditional way of compiling code is extensively redesigned for other purpos-
to generate machine code directly for the tar- es, it might be beneficial to redesign it to
get machine. However, the APZ 212 40 will run on a new platform using a commer-
not follow this approach, since it is not com- cial development environment and com-
patible with assembler additions to legacy mercial implementation languages. An
applications. Instead, the PLEX code will be example where this approach has been
compiled to assembler instructions. This pursued is from TDMA, where existing
makes the incorporation of existing assem- MSC functionality in AXE, which han-
bler additions transparent. The gains in han- dles radio network control (RNC), is
dling will far outweigh the minor loss in being redesigned on the NSP.
performance. • accessing new functionality—instead of
developing new functionality in AXE, it
might be more cost-effective to combine
Communication services it with functionality already available in
The application modularity architecture another platform. For example, the next-
was introduced to define the AXE 106 sys- generation switch (ENGINE) will pro-
tem (Figure 10). The basic principle was to vide ATM capabilities to a transit node
divide the AXE system into different types by combining AXE with AXD, thus
of system module: application modules using ATM functionality supported in
(AM), resource module platform (RMP), ex- the AXD 301.
isting source system (XSS) and the APZ • reuse—instead of redeveloping all exist-
computing platform. ing AXE functionality on a new platform
An important feature of this architecture it might be more cost-effective to com-
is that all inter-application-module com- bine some of it with new functionality on
munication takes place indirectly via com- another platform. An example from
munication services within the RMP (APC, TDMA is where existing maintenance
CLMT, and OBMAN function blocks). The functionality of the old radio network is
logical names of protocols and services are kept in AXE while control functions are
used at the application level, which the moved to the NSP.
RMP resolves to physical addresses. This is To support cases that involve distributed ex-
similar in concept to an object request bro- ecution of traffic AMs, new communication
ker (ORB). services will have to be introduced, since
present-day communication services indi-
New kind of openness rectly assume that the AMs execute on the
With this architecture, one or more of the same platform. For instance, no error cases
application modules can, in principle, exe- are handled if the other platform is not avail-
cute on a physically separate platform (AXE able or if the communication link between
or other), since the physical distribution is platforms fails.
hidden by the RMP. This possibility has al-
ready been exploited for charging informa- New communication services
tion and statistics; that is, the FOS and STS
have been moved to the AP platform. This TRH
points to a new kind of openness in AXE The transaction record handler (TRH) pro-
that can be used for tocol—which was developed to support the
• boosting capacity—some application communication needs of the control inter-
modules require huge amounts of capac- face between AXE and AXD 301 in

Ericsson Review No. 1, 2000 39


ENGINE—adds a proprietary layer on top ed on the IPNA, in order to offload the cen-
Other of TCP/IP and runs over the STOC hardware. tral processor.
Object function block

CORBA Summary of communication services


OBMAN Wrapper The common object request broker archi- The PLEX interfaces to different communi-
tecture (a standard promoted by the Object cation services (Figure 12) are similar and
Management Group, OMG) is frequently typically provide support for
CORBA
communication
used for O&M. Several prototypes have been • defining the messages in a protocol (used
developed to test its usability. One proto- to support byte ordering and transport
OCITS-1
type implements a CORBA communication format of data);
OCITS service (a PLEX-ORB). Wrappers that pro- • setting up communication connections;
CP
vide service interfaces within AXE to ap- • sending and receiving messages (simple
TCP/IP
protocol plications executing on other platforms are data, communication buffers and message
stack also implemented (Figure 11). The wrap- buffers);
pers use the PLEX-ORB for communica- • tearing down connections;
STOC (RPG) tion. The interface between platforms has • reporting the failure of communication
Ethernet been specified using the interface descrip- peers and handling diverse AXE recovery
tion language (IDL) defined by OMG. scenarios (Forlopp, minor and major
This implementation proves that O&M restart); and
Figure 11
Implementation structure of the CORBA
functionality can be moved to another plat- • handling duplicated links and, if a failure
prototype in AXE. form, using IIOP to communicate with the occurs, switching from the failing link to
traffic applications that remain in AXE. As the working link.
an added advantage, it is easy to provide a
Web-based O&M interface to applications
that use an object request broker.
Conclusion
The adjunct processor platform is chiefly
CSH made up of sourced components. Ericsson
A main drawback of distributing a traffic provides additional value to the adjunct
application over more than one platform is processor platform in AXE by adding sup-
that a lot of CP capacity is required to send port for supervision, alarm handling, event
messages and data between the platforms. A handling, upgrading software, audit log,
third alternative being developed, called the and access authorization, as well as support
connection service handler (CSH), aims to for interfacing the CP. Since the availabili-
minimize CP load during communication ty and robustness requirements for the AP
from AXE. are less severe than for the traffic platform,
The CHS communication service will be Ericsson can source components compara-
based on the proprietary inter-processor tively high in the layered component model.
communication (IPC) protocol used within The RPP and Ethernet packet switch
an NSP cluster. The IPC can run on raw Eth- board are composed of several standard com-
ernet (and on top of other standard proto- ponents. Ericsson has also, among other
cols, such as the user datagram protocol, things, added interfaces to AXE, and O&M
UDP, or ATM) with low overhead; the IPC functions according to the AXE standard.
header size is only 7 Words compared to 16 Because the RPP and EPSB support indus-
Words in TCP/IP. Moreover, IPC is robust try-standard interfaces, other commercial
and supports efficient addressing. In its first products can easily be introduced as they be-
release, the IPC protocol will be terminat- come needed.

TRADEMARKS

Diskkeeper is a registered trademark of


Executive Software.
Pentium is a registered trademark of Intel Cor-
poration.
Sun Enterprise Volume Manager is a registered
trademark owned by Sun Microsystems Inc. in
the United States and other countries.
Windows NT is a registered trademark of
Microsoft Corporation.
WinZip is a registered trademark of Nico Mak
Computing, Inc.

40 Ericsson Review No. 1, 2000


The interplatform network adapter will
add the industry-standard 100 Mbit/s Eth- AXE10
ernet interface to AXE CP, thereby im-
proving reload and backup times and pro- APZ/APT applications
viding support for interworking with traf-
fic applications on other platforms. The IPN APSI
adapter is composed of several commercial COMS
components, such as processor chips, oper- CORBA TRH CSH
CP communication RMP
ating system, and protocol stacks.
Ericsson will introduce several commer-
cial products, interfaces, and key software OCS
OCITS-1
components into the next generation APZ. OCP
In-house hardware and software compo-
nents will also be introduced to enable the
next-generation CPU to execute legacy ap- IPNA IPNA
RPH-S RPH-S
plications, to adhere to ISP requirements, RPH IIOP
level TCP/IP IPC
and to interface legacy parts of the AXE sys-
tem. Future technology that can be intro-
duced into AXE includes larger symmetric
STOC
multiprocessor (SMP) clusters and super- RP
level IIOP
scalability. TCP/IP
The introduction of new communication
services in AXE will make it possible to
combine the functionality in AXE with that 100 Mbit/s Ethernet RPB PMC 10 Mbit/s Ethernet 100 Mbit/s Ethernet
of other platforms, in order to support step-
by-step migration and reuse. For example, Figure 12
the transaction record handler has been Overview of the communication services in AXE.
shown to work for ENGINE traffic applica-
tions, the IIOP works with O&M functions,
and the connection service handler is being
developed to work with traffic applications.
The communication services described will
either run on raw Ethernet or add function-
ality on top of TCP/IP or IIOP.
During the past several years, Ericsson’s
hardware and software developers have been
using commercial components to build the
AXE system. This process has been accel-
erated and, as this article shows, commer-
cial products and interfaces will soon make
up the core components of AXE. This
means that Ericsson can fully focus its de-
velopment efforts on providing telecom-
munications characteristics, such as high
REFERENCES
availability, robustness, fault tolerance, and
capacity.
1 Ericsson, M. and Koria, N.: Adjunct proces-
sor—A new AXE-integrated open-standard
computer system for call-related data
processor. Ericsson Review Vol.74 (1997): 2,
pp. 82-90.
2 Granbohm, H. and Wiklund J.: GPRS—Gen-
eral packet radio service. Ericsson Review
Vol. 76 (1999): 2, pp 82-88.
3 Blau, S. and Rooth, J.: AXD 301—A new gen-
eration ATM switching system. Ericsson
Review Vol. 75 (1998):1, pp.10-18.
4 Hennert, L. and Larruy, A.: TelORB—The
distributed communicataions operating sys-
tem. Ericsson Review Vol. 76 (1999): 3,
pp.156-167.

Ericsson Review No. 1, 2000 41

You might also like