|
|
 |
|
|
|
|
|
| The Itanium processor–based
NonStop server will deliver cost
savings and greater flexibility in
addition to a significant price/
performance boost, according to
Steele. “In the current NonStop
server environment, development
systems have the same level of fault
tolerance as production systems,”he explained. “But with the new
Itanium processor–based system, you
can drive the cost down by using a
non-fault-tolerant simplex machine
for development. For production, if
you just need the same kind of fault
tolerance you have today, you
deploy a duplex system. And if you
have an extremely critical application,
you can make it a triplex system;
HP hasn’t even calculated the number
of ‘nines’ that you can get from one
of those.” |
| |
|
|
|
| |
|
|
| This is not to say that Sprint doesn’t
make extensive use of current NonStop
technology; in fact, it’s a relationship
that has grown stronger over the years.
“The NonStop platform has been at
Sprint since the very beginning,”
Steele stated. When Sprint was just a
long-distance and local voice network,
the company needed processors at
the switch site that could offload
some of the data collection and
processing tasks. NonStop server–based SCPs were also deployed on
the long-distance side of the network
to handle some of the more complicated
routing algorithms that a
long-distance switch couldn’t do.
Sprint put a NonStop system in
place to collect and save operational
measurements and billing information
from the switches. Some of the
company’s enhanced services—calling cards, operator services,
and hospitality services—run on a
variant of the NonStop server–based
SCP. Provisioning for the long-distance
switches is handled by a
NonStop system.
In Steele’s view, one of the best
things about the NonStop platform
is its low total cost of ownership
(TCO). “As you move into the second
and third year, far fewer people are
needed to operate and maintain a
NonStop system than other high-end
computing platforms,” he said.
“A single person operates our SCPs,
which consist of four systems; if you
look at some of the other brands of
SCPs, it can take as many as nine
people to keep them running. And
the support for open standards is
another area in which we save
money, because we don’t need to
invest in special programmers. For
developers on the OSS side of the
NonStop system environment, it’s
just like working on any other
UNIX® platform.”
Sprint takes advantage of other
HP products and services as well.
“We’ve got HP platforms throughout
the company, both in the network
and supporting the network,” said
Steele. “In addition, Sprint recently
switched to HP OpenView as the
enterprisewide internal operations
platform.” Over the past few years,
Sprint has consolidated operations
into two groups: network and
enterprise. OpenView was chosen
as the platform to manage all the
computing systems. The NonStop
servers will be incorporated into
this overall management framework
in the near future. |
| |
| | |
|
|
|
A pervasive presence
| |
|
|
|
Sprint relies on NonStop servers
in several key areas:
-
Service control point (SCP) system
comprises four network nodes across
the United States and handles some
3.7 billion transactions per month.
- Service Delivery Platform (SDP)
produces transaction processing
applications that insert call processing
data into live network elements in
near real time.
- Sprint Call Center handles station-to-station and person-to-person calls,
international general assistance and
directory assistance, fraud features,
and other services including call
screening and routing.
- E-911 handles emergency services.
- ODS/ZLE is Sprint’s name for the
operational data store (ODS) on
HP’s ZLE architecture. Sprint call
detail record (CDR) and operational
measurement (OM) volume is approximately
356.5 million per day on eight
ODS/ZLE platforms.
|
| |
| | |
|
|
|
|
| |
|
|
|
HP’s Zero Latency Enterprise (ZLE)
framework made its first appearance
at Sprint about the same time
the year 2000 rolled in; in fact, the
two events were inextricably linked.
“Faced with Y2K issues and a huge
increase in data coming into the
network, we put a challenge out to
the industry,” recounted Steele. “We
said: ‘Tell us how you would solve
the problem of capturing and storing
20,000 records a second, with geographical
replication so that we
never lose connectivity and we
never lose a record.’ At that time, HP
(then Tandem) was the only company
that came back and said: ‘We can
do that.’ That collaboration with
Sprint was a foundational element of
what became the ZLE architecture.”
Many benefits have accrued from
Sprint’s implementation of the ZLE
architecture, but customer service
tops the list in Steele’s mind. “The
ZLE framework gives us a way to
see exactly how traffic is traveling
through the network,” he stated. “It
enables us to troubleshoot problems
and optimize service in real
time for our customers.”
A network the size of Sprint’s is
very dynamic due to the proliferation
of temporary “virtual” circuits
such as Frame Relay and Internet
Protocol links. In the event of customer
problems or network errors,
the ability to reconstruct the state of
the network is critical. Sprint keeps
an online topology of the network
on its ZLE framework–based system,
making it possible to view the state
of the network at any given point in
time; tie it back to particular call
detail records, faults, and alarms;
or help trace a problem that a
customer may have experienced.
A duplicate ZLE architecture–based system resides in a second
data center. HP NonStop Remote
Database Facility (NonStop RDF)
software moves the data between the
two systems. At one point, Sprint
considered using a storage area
network (SAN) configuration for
replication, but that approach was
deemed inadequate for missioncritical
data. "The SAN could not
synchronously replicate the data
across the geographical distance of
1,300 miles, so there was a period of
time during which data could be
lost,¡± noted Steele. "But with the
NonStop system and NonStop RDF,
the data is protected all the way
through the flight. Until it's written
and stored in the other database, the
initial transaction is not closed. So
we don't risk losing the data at all." |
| |
| | |
|
|
|
|
|
|
|
