It's all happening. The race is on. Dominance is sought in the ISP
(internet service provider) space.
Sun reckons some 30 per cent, and growing, of its E10000 server
business is internet activity, including mobile companies and
wireless internet access, while 'all the internet banks use Sun -
Egg and Smile run on E10000s, for example'. Services running
through lastminute.com use a farm of 100 Sun boxes.
The internet business productivity purchasing service works.com
has two IBM RS/6000 S80s running under Aix at the core of its
internet operations. The S80s power the company's web site, which
provides small and medium-sized businesses with discounts, and fast
deliveries on 20,000-plus business products.
Network Solutions has selected an RS/6000 S80 to power the
internet's master computer for all web addresses, and the global
constellation of name servers for .com, .net, and .org. NSI also
provides domain name registration services for all available
country code top-level domains, eg .de (Germany), .fr (France) and
.uk (UK). NSI registers most web addresses worldwide through
various channels, including 260-plus companies in 30-plus countries
in its premier programme, and 41,000-plus companies in the
affiliate programme.
Now IBM has launched IBM ParaBlue for ISPs - 'a flexible
architecture for ISP and portal platforms running in a Unix
environment'. It's part of a portfolio of offerings for ISPs in
Unix and Linux environments, covering storage, web servers, and web
management software:
Big Blue's storage division has released two enterprise storage
servers (ESSs), 'a step ahead', enabling ISPs to scale high growth
in demand, and 'storage area network (San)-in-a-box', to deliver a
pre-packaged combo of hardware, software, and expertise to get ISPs
started. The San-in-a-box is based on fibre channel attached to an
ESS, San fibre channel switch, and IBM Global Services installation
and implementation services.
Servers are based on the latest generation of RS/6000 Unix mid
range boxes - models F80, H80, and M80, are based on the same
64-bit copper technology of the S80 top-of-the-range. A range of
packaging options is available for these servers.
Tivoli's service provider solutions constitute the final
component, and enable the management of ISPs' infrastructure and
subscribers.
John Lutz, IBM Net Generation Europe, says: 'ParaBlue for ISP
enables ISPs to deploy integrated end-to-end service delivery
management across back office IT operations, data, and voice
operating networks, and wireline and wireless networks, allowing
them to manage and deliver communications or information services
to customers.'
The mantra out of Sun that 'the network is the computer' has
been pursued for 18 years. Not just said, but implemented, in terms
of both product development and business strategy. On products, the
company has come out with the Netra t1 1RU system - designed with
input from industry key players - which, with the E420R, became
Sun's highest unit selling server within its first full quarter of
business in 2000.
David Allinson, Sun's UK network server product manager, says:
'In this same quarter, early 2000, Sun as the market leader of Unix
servers increased its share to 34.3 per cent, from 1999's 28 per
cent. Share was mainly captured from Compaq and HP, which combined
had a 1999 share of 49.9 per cent, which dropped to 38.1 per cent
in Q1 2000, with Compaq securing the number two position from
HP.'
Growing opinion has it the service-driven network will become
the method of delivery for IT or network services. Although it's
forecast the number of PCs sold and used to access these services
will grow steadily (says IDC), by far the greatest growth to
outstrip the existing PC interface will stem from a range of other
devices. These include Wap phones, digital TVs with web access, web
access pads, internet access devices, and consoles in public
places. All will access an increasing number of services.
'Windows will then become a far less important consideration' is
the Sun decree (one can sense Scott McNealy smiling in the
background). The growth will be in the connectivity, bandwidth,
processing, and storage capacity of the network. Sun's 'three big
bets' as the blueprint for the development of the service-driven
network are:
Massive scaleability - systems will be required to scale up
where data cannot be easily replicated. Here, clustering becomes an
important factor for reliable access to large active databases.
Continuous real-time computing is essential for a 24x7x360
network economy. People will be continuously connected to the
internet, no matter where they are, and will want instant access to
information, comms, and other services at all times. The days are
gone of 'it's the weekend, so let's do the maintenance'. The
challenge now is to design software, build systems, and provide
support services that eliminate downtime, and deliver real-time
responsiveness.
The integrated stack. With massive systems scaling, and
continuous real-time services, the back-end system will have to be
an integrated hardware and software stack. The microprocessors,
storage, system software and middleware will all need to be
seamlessly integrated for dependable operation.
So how fares the noble Unix? Swedish start-up Bluetail, which
grew out of Ericsson's research labs in Sweden, develops and sells
Unix based software to internet organisations. The company
specialises in systems that maximise efficiency in the internet
operating environment.
Being used to the Unix platform, Bluetail came under pressure to
support Microsoft. While it offers support for both MS2000 and
Unix, Bluetail prefers Unix for its application programming
interface, reduced costs for ISPs using the solution, reliability,
portability across multiple vendor platforms, and less maintenance.
It has launched a solution that claims to make web site crashes a
thing of the past.
Running on a PC, the 'ISP robustifier' is a software based
solution which, as well as providing load-balancing, so one server
is not overloaded while others are idle, also keeps the system
working, even if one of the servers crashes. The workload is
distributed across the remaining servers. Another feature is a
function preventing new users accessing the system if the servers
are close to overload - making site surfing much faster for
existing users.
Three solutions are offered: mail robustifier for e-mail
reliability; web prioritiser for internet site traffic management;
and directory robustifier offering faster access to web sites.
A fast-responding web site is generally considered to be key to
building a successful online presence. Yet according to research
conducted by IDC and commissioned by Amdahl, few businesses know
how responsive their own web sites are - let alone how long
potential customers might be prepared to wait for the site to
download.
Charlie Abrahams, Amdahl UK md says: 'Valuable business is
likely to be lost because of this. Bear in mind competitors are
only a click away. Still it's amazing companies don't know how long
their customers are waiting to use the web site.' l
Managing to deliver
There's scant discussion about what ISPs must do to win in the
fast-moving IP-centric (tele)comms market. They who deliver the
services customers want, when they're wanted, with the required
quality and security, and at an acceptable cost will win. Having
the fastest Unix server farm will not guarantee success. The
challenges facing ISPs called upon to deliver multiple services -
data, text, video, voice, etc - over a converged IP-centric
infrastructure are manifold. How does the ISP switch on such
services across an increasing number of network devices? How are
different devices from many manufacturers contained under one
management system? How so quickly and accurately? How can it be
ensured things like security, connectivity, etc are activated as
intended?
Enter policy based network management (PBNM). OK, this isn't
likely to appear on users' networks tomorrow. Today, most would
probably add on more bandwidth to support more and more users and
applications. Others might say 'stuff it - so let the e-mails fall
off the edge of the net'. PBNM systems can have an architecture
devised by the DMTF (distributed management task force) and IETF
(internet engineering task force). What? Policy decision points
(PDPs) receive information from the policy repository and make
decisions based on those policies (eg the e-mails get priority).
Policy enforcement points (PEPs) - eg gateways, switches, routers,
firewalls - implement the decisions passed on by the PDPs. Sounds
simple, but it isn't. Current activity is homed in on QoS and
security.