|
|
article page | 1 |
2 | 3 |
4 |
5 |
6 |
7 |
8 | 9 |
Maintainability,
Manageability, Scalability, Accountability,
and Durability are directly contained
in the concept – sort of the
faces of one six-sided dice. Too
these, Securability
and Survivability are recent critically
expected features. Just because a
single KPI, the notorious five-nines,
is the flagship indicator, does not
exclude the specifications for these
other factors. This is very evident
in the Bellcore documents which,
in addition to all these, link in
Safety. With software and servers,
Reliability, Availability, Serviceability,
Usability, and Install-ability combine
in the acronym RASUI.
Getting Realistic: Big
Customer Requirements
The US Government
has issued a comprehensive RFP [Networx
Universal RFP; TQC-JTB-05-0001; 05/06/2005]
for bidders wishing to supply the
next generation network for the US
Government. Considerable thought
has gone into setting acceptable
quality levels for networks – taking
into account the differing use needs
and technical facilities of each
network technology. This is perhaps
the most complete, openly viewable
and usable description of realistic
network service goals. Many dozens
of tables describe the acceptable
availability and other quality measures
for each network technology. Further
they go a step further and provide
for both normal and mission critical
network needs. “For certain
services, when required by agency
customers, two service levels are
specified. Routine service levels
apply for most Government applications.
Critical service levels are defined
for Agency applications requiring
higher levels of availability, performance,
or restoral <sic>criteria.” Voice
networks are generally given normal
availability goals of 99.5% and critical
service goals of 99.95%. Data network
technologies are usually about one
magnitude more stringent – 99.95%
for normal use and 99.995% for critical
service uses. Networx provides a
modern and realistic reliability
guideline for networks. It is the
new baseline benchmark, below which
a network is unacceptable, and above
which is the narrow window of competitive
differentiation.
Nevertheless, SLA’s
are subject to extreme competitive
pressures, which have backed their
availability promises right up against
the wall. “Verizon Business’s
Network Availability Commitment is
to have its IP Network available
100% of the time … Verizon
Business will credit Customer’s
account if Verizon Business fails
to meet this Network Availability
Commitment during any given calendar
month.” [‘Verizon business
SLA for DSl.pdf’] Since 100%
availability is very difficult to
achieve, SLAs become a kind of amortization
game. You give an SLA with the up
front knowledge that you will inevitably
give back
|
|
some revenue
to your customers. Access
links from the customer
premise into the network
have always been the achilies
heal when engineering for
redundancy. Even with low
failure rate, a MTTR of
about 3 hours on access
links forces availabilities
that are likely far from
100%. Instead, the revenue
loss from SLA penalty payments
is offset by increased
prices or customer satisfaction
with the SLA. In the early
days SLAs actually made
money. One prominent carrier
offered the only SLA and
charged a premium because
their ‘network was
so good it could support
an SLA.’ While truthfully,
their network was no better
than most other carriers,
they earned 10x more in
extra income from the higher
prices charged than they
had to payout for SLA violation
rebates.
But
today is perhaps different:
claims of network performance
can be validated more
easily by customers. “Because
of its built-in redundancy,
Internap can offer aggressive
performance guarantees,
including 100% availability,
0.3% packet loss and
45 milliseconds of latency. “We
offer 100% availability
border to border, from
where a customer intersects
our network to where
the customer's traffic
leaves our network," Flynn
says.” [Case Study
submitted by Carolyn
Duffy Marsan of Network
World to TMF]
Getting
better and better
Several organizations
are devoted to helping
make networks and network
components better. For
example, there is the Technical
Committee on Communications
Quality & Reliability
(CQR).
It focuses on and advocates
worldwide communications
and reliability on behalf
of, and within, the Communications
Society (formerly known
as the Quality Assurance
Management Committee (QAMC)
). And while the customer
driven goals may be less
than five-nines, this level
of performance is realistic
and achievable across complex
networks when best design
practices are utilized.
The fine-nines, never a
requirement myth for equipment,
is now moving from myth
to reality for networks
too.
And software, once considered
hopelessly incapable of
quality, now sets itself
some of the highest reliability
goals. CGL (Linux) OSDL
standard allows system
registration to 'carrier
grade' standards. “Carrier
Grade Linux' is a set of
specifications which detail
standards of availability, scalability, manageability,
and service
response characteristics which
must be met in order for Linux to
be considered "carrier-grade" (i.e.
ready for use within the telecommunications industry).
Improvements are expected.
. “Carrier-grade
is a term for public network
telecommunications products
that require up to 5 or
6 nines (or 99.999 to 99.9999
percent) reliability….
The term "5 nines" is
article
page | 1 | 2 | 3 | 4 | 5 | 6 | 7
| 8 | 9 |
|
|
|
|
