Computer Weekly reviews how the main server makers are
formulating strategy for the datacentre, and whether any viable
alternatives are on the way. Our conclusion: keep an eye on
Dell
As the only original
datacentre player still in the top league,
IBM has impressed analysts with its strategy. When the company
recently previewed its latest-generation System Z mainframe, the
Z10, it was the first since the release of the Z9 three years
ago.
"There is a general recognition of the value of mainframe and
enterprise servers," says Roy Illsley, senior research analyst at
the Butler Group, and one of the experts who was briefed on this
new plank in IBM's datacentre technology. "The mainframe can still
be at the centre of IBM's datacentre strategy, but the more fluid
parts of its IT offering - the enterprise servers and the X86
servers - can be deployed as commodities where needed."
The new IBM mainframes are a bid to keep a customer base that
spends millions of pounds on new systems, loves the old-fashioned
resilience and trustworthiness of big Iron, but continue to be
tempted by the cheaper, more flexible alternatives.
Mainframes do not crash and are boringly reliable - claims no
one could make about servers. And one of IBM's four socket Intel
servers costs a fraction of the price, runs faster than the
mainframe and comes with cheaper software.
IBM user groups continue to worry about the conflict between
distributed systems and mainframes. In a recent survey by IBM user
group Share, 23% admitted they were planning to increase usage of
their mainframes, and 19% said they would cut back on mainframe
reliance or say goodbye to the systems completely.
Since Z9 came out in 2005, IBM has continually tinkered with the
platform, making a series of announcements of gradually improved
administration. It spent £50m on simplifying the mainframe
environment to encourage use of the systems in service-oriented
architectures (SOA). But one problem remains - not many young IT
experts even know what a mainframe is, let alone how to work with
one. To solve this difficulty, IBM is funding training programmes
at universities to create a base of available mainframe-trained
talent.
Arguably, HP has surpassed IBM as the biggest computer company
in the world. Although it prides itself on its engineering culture,
HP seems to be heavily reliant on marketing to win attention. The
phrase "managing risk" is bandied about a lot in its presentations,
possibly in the hope that it will resonate with buyers.
The message from HP is that it wants to deliver more efficiently
used assets. "Everything has to be saved now," says Phil Dodsworth,
director of HP Datacentre Solutions. That means energy efficiency,
power and cooling and lowering operating costs, he adds. "Our
strategy is all about managing risk for our clients, and we manage
risks through these processes. It's not a silver bullet."
HP sees
virtualisation as an important strand in what it wants to
deliver - but that is a long way from the full story. "Every other
supplier seems to imagine that virtualisation is a cure all for
everything," says Dodsworth, "but some applications don't lend
themselves to virtualisation at all."
HP's strategy is very much cost-driven rather than
technology-led. "That's what our customers have been telling us,"
says Dodsworth. "It's the cost of the datacentre that is crippling
companies, not the strategic advantages that technology can
deliver." A case in point is
Barclays Bank's new datacentre in Gloucester, where the biggest
selling point - smart cooling - is nothing to do with techniques
for processing data, but all about power management and keeping
bills down.
Illsley says, "HP is engineering-led and one of the biggest
computer manufacturers in the world, so it has the capability in
our labs to tackle any issue. The fact that they look at energy
efficiency, expressed as dynamic smart cooling, intelligent air
conditioning and blade designs, tells you what the priorities are
these days for HP."
At HP, the acronyms bandied about now do not relate to
applications or processing engines - they are more likely to be
something like PUE (power usage efficiency).
Dodsworth comments, "Once we measured IT performance by MIPs -
now the benchmark is in watts."
Sun's strategy, meanwhile, is best described as a datacentre in
a box. The company offers a preconfigured, self-contained
datacentre that gives an instant maximum of 1,200 servers and 3.5
petabytes of storage in a network. Sun says it cuts your roll-out
time and your electricity bills.
Sun's 20ft Project Blackbox comprises racks of servers and
storage in a shipping container. This model is still fairly novel,
but could enable companies of all sizes to move their datacentres
to the best source of power. There are fewer and fewer places in
big cities that have a licence to draw that much power from the
national grid, not to mention good bandwidth connections and
cooling capacity.
Analysts like the idea, though. Illsley observes, "Big
corporations are predominantly based in London, but there's hardly
anywhere within the M25 that can cope with the massive drain on the
national grid, and then they have to find sufficient floorspace at
the right price. There is certainly nowhere like that in the City,
so
datacentres have to migrate to wherever the capacity is right
and there are cooling facilities. In the UK, that could be
Scotland."
Sun's system's agility could win it many orders one day. When it
has an installation in the UK, we will find out whether its claims
for plug-and-play technology stand up. Until then, the idea of an
out-of-the-box plug-and-play datacentre sounds rather fanciful.
Sun still offers users the ability to mix and match its servers
with other platforms.
Meanwhile, Fujitsu Siemens Computers (FSC) recently unveiled the
latest version of its Primergy servers, which it describes as
energy-savers for the datacentre.
Like all other suppliers, FSC's strategy for the datacentre
primarily concerns energy - giving end-users what they want. But
the scalability of its range has impressed analysts, such as
Gartner's datacentre analyst, Rakesh Kumar, who insists only
newer-generation server technology can provide the flexibility to
turn the datacentre into a "living organism".
"New
datacentre designs need to be based on flexibility and high
levels of monitoring," says Kumar. "They should be perceived less
as a static structure and more as an agile living organism that
evolves as the server and storage infrastructure changes." He says
FSC is doing as much as anyone to make that possible.
The servers use quad-core processors to reduce the load on each
chip, in an attempt to bring them up to mainframe reliability
levels. Meanwhile, its 2.5in discs use one-fifth less power than
traditional 3.5in drives. The only departure from mainframe
technology - apart from flexibility - is that the new units came
out with Windows Server 2008.
Dieter Herzog, executive vice-president of infrastructure
products at FSC, says the company's datacentre strategy around its
new Primergy servers dovetails with the new Windows Server 2008.
"It's the best way to get higher efficiency, better quality of
service and increased agility of IT operations," he adds.
FSC's latest marketing strategy warns IT professionals: "Dump
your old server." Its campaign urges companies to upgrade their
network infrastructure.
This could reflect its position as something of an also-ran in
the server market. Some argue Dell should be in our top four, but
FSC retains a high ranking because it is heavily focused on
standards, which protects end-users from supplier lock-in. Wasn't
that what caused all the trouble for datacentres originally?
According to IDC, Dell was the number three server maker
globally in 2007. Not surprisingly, green IT is one of the main
factors driving Dell's datacentre computing strategy, says Dr
Albert Esser, the company's vice-president of datacentre
infrastructure.
"When it comes to ageing datacentres, customers have two
options," says Esser. "Neither is good for the environment. They
can rip and replace the entire infrastructure or build new
facilities."
Dell's strategy is to cater for the uncertainty many investors
will experience in the current economic climate. It is gambling
that customers will want to be energy efficient by prolonging the
life of their datacentres - not replacing them.
Esser talks of revealing a hidden datacentre - within the
existing facility - to clients. Dell is banking on the fact that
virtualisation will boost overall utilisation and remove the need
for current levels of physical computing resources.
Dell's strategy, says Esser, is based on a trinity of issues
that drive datacentre computing - virtualisation, consolidation and
grid computing. "CIOs should standardise equipment, consolidate
servers, storage and applications and automate management. Then you
will see a rapid return on grid computing."
Dell has partnered the likes of Oracle, EMC and Intel to create
Project MegaGrid, which tries to formulate a single, validated set
of deployment best practices for grid computing.
Among the alternatives, Unisys is troubled by financial
problems, and its strategy is to shift its servers and storage
business from being hardware-centric to an approach encompassing
more influences, such as software and services, including
virtualisation. Unisys is upgrading its Intel-based server line by
adding quad-core processors to its ES3000 servers and releasing an
ES5000 series of blade servers.
That other hardware stalwart, Bull, has built a strategy around
simplifying the complexities of virtualisation. Its all-in-one
virtualisation server was launched in March, aimed at companies
that may be daunted by the slog of cutting the number of deployed
servers. Its NovaScale VMBox is currently available in two
configurations and supports VMware ESX Server 3.5, with the VMware
VirtualCenter Foundation management platform hosted on a separate
NovaScale server.
But how can Bull simplify virtualisation? Isn't it meant to be
complicated?
Its idea is to pre-integrate everything. Along with the servers,
it is including storage subsystems, pre-loaded software and Bull's
Calypso backup software. These configurations are pre-integrated,
which Bull claims enables "faster deployments of infrastructure at
a lower cost".