Quite a list Douglas.
HP touchy about the Tachyon chip?
..........
HP disputes Test Center;
To The Editor: CRN is to be commended for its in-depth look at the future of
Fibre Channel in mass storage. However, I am writing to express concern over
a comment that was made regarding the Hewlett-Packard Tachyon 2.1 Fibre
Channel protocol controller IC in the "Lab Analysis" section of the "Enterprise
Storage" roundup (CRN, Jan. 4).
First, CRN [Test Center] wrote that the HP Tachyon 2.1 protocol control
engine "could be problematic" because it was found in the systems that
performed the slowest in your tests. HP believes that this comment is
misleading and unfair to the HP Tachyon product family.
Overall storage-system performance is the function of many components,
including the host computer and operating system, host-adapter card and its
device-driver software, architecture and firmware of the storage system and
disk drives. The interaction of all these components-and not the action of one
specific component-determines overall storage-system performance.
Second, CRN wrote that the HP Tachyon 2.1 was developed "prior to the
standardization of Fibre Channel," which could be interpreted to mean that HP
Tachyon 2.1 is not compliant with Fibre Channel standards.
The HP Tachyon 2.1 is fully compliant with FC-AL, FC-PH and other Fibre
Channel standards. The fact that it was developed in parallel with the
development of FC-AL and FC-PH in no way diminishes its ability to deliver
fully compliant, enterprise-class performance and features to the many different
applications in which it is designed-in.
HP has dozens of OEMs that have selected it for use in their products, including
EMC, Clariion, Compaq, Hitachi and Ciprico-the winner of the roundup-which
uses Tachyon in its Fibre Channel RAID storage subsystems.
HP remains fully committed to providing OEMs and distributors with
high-quality, high-performance, Fibre Channel I/O products.
Its second-generation Tachyon TL and Tachyon TS controller ICs, as well as
the host-adapter product line based on these controllers, are evidence of this.
The recent performance record of over 31,000 input/outputs (I/Os) per second
by the HP HHBA-5100A host-adapter card is proof that the HP Tachyon
family is capable of industry-leading performance. I urge your readers to
evaluate HP's Tachyon family of Fibre Channel products in their specific
applications before making conclusions.
Rich Lautzenheiser
Strategic marketing manager
Fibre Channel Solutions Group
Hewlett-Packard Co.
Fort Collins, Colo.
---
Centralization......
techweb.com
Centralization's New Twist -- Corporate IT takes center
stage as companies use the Web to recentralize
Lenny Liebmann
Fueled by high costs and incompatible systems, IT managers are shining up their
dancing shoes and reasserting their role over technology across departments
and lines of business.
In fact, many Fortune 1000 IT departments are taking center stage, deploying
thin clients, server farms and storage area networks-technology that clearly puts
them back in the corporate spotlight. At first blush, it would be easy to view the
new IT architecture as a return to the old days of mainframe computing-with
browsers taking the place of 3270 clients, and IP as the new universal network
protocol instead of SNA.
But there's a new twist this time around. True, management of systems and
network infrastructure is back under the control of enterprise architects and
administrators, but the glass-house attitude and monolithic approach are gone.
Web technologies let IT departments logically centralize infrastructure
management without necessarily centralizing it physically.
"Our manager in charge of enterprise messaging is in Concord, Calif., and his
direct report in charge of Notes is in Jacksonville, Fla.," explains J. Kevin
McLees, vice president of Bank of America's Strategic Technology Group, who
works out of an office in Richmond, Va. "The sophistication of the network
means that we all don't have to be in the same place to do our jobs,'' he adds,
pointing out that the high bandwidth and transparency of most corporate WANs
makes location a nonissue.
For users, Web technologies mean that maintaining local possession of
computing resources in their business units-a sticking point when PCs first came
on the scene-is far less important. "One of the key things the Internet has
demonstrated to the business world is that even if a server is located on the
other side of the world, you can still get to the information," says McLees. "So
now it's easier to talk to a business client about keeping servers someplace
where they can get the best care and feeding."
And that care and feeding is best done from a central location. "The perception
for users is that they're going to get a very easy-to-use interface that connects
them to the mainframe," explains Toby Younis, vice president for the
consultancy Meta Group's Executive Council. "The fact that the mainframe is
really a bunch of physically disparate information sources is really irrelevant to
the user."
Cost is driving the move to central IT as much as anything. For example, why
have high-priced technicians perform routine housekeeping tasks such as
backup on their local resources? It makes more sense-and it's cheaper-to have
specialists perform those tasks remotely across the enterprise.
"Centralization is driven by a price/performance argument," says Hal Uygur,
vice president for enterprise systems operations at investment bank Goldman,
Sachs & Co. "A 24-by-7 support organization with less expensive human
resources can deliver support that frees up more expensive resources, like
systems administrators and engineers, to focus on more value-added work."
According to Uygur, tasks such as automatically monitoring failures should be
done by centralized "operators" who are less expensive to hire and train than
engineering types who know all the ins and outs of operating systems and
network protocols. An experienced Unix expert, for example, can cost 60
percent or more than a recently certified network administrator.
Reliability is another concern that resource consolidation addresses. "When you
decrease the number of devices you have on the network, you increase your
aggregate uptime," says McLees. "If you have hundreds of piece-parts for a
service like messaging and they're already pretty fragile as it is, your uptime is
never going to reach 99-point-anything."
Both reliability and economy are improved by centralization's cousin:
standardization.
"If you want an IP address from us, then we request that you have the right
version of NT, the right version of Office, the right amount of backup, and so
on," says Edward Tunstall, information officer for Enterprise Information
Systems at pharmaceutical giant Eli Lilly & Co. "That's how you make sure that
anyone in the organization can share information and that your unit costs for PC
support stay low."
Centralization also has become critical to supporting business growth.
"If we want to put a Unix server in a new location, we don't want to have to
rely on having a local systems administrator there as well," says Uygur.
"Instead, you want to be able to rely on your central administration and backup
team to handle it."
Meta Group's Younis says that because companies typically grow through
acquisition and short-term partnerships, the ability to create logical centralization
without physical unity is essential. "You can't quickly create these business
relationships if you still think in terms of bringing everything into one place in
order to share it," he says.
The "virtually centralized" management that the Web permits is becoming a key
for global companies that need to keep the enterprise network humming and
can't afford a full, three-shift technical staff at company headquarters (see
"Centralization: What's The Internet's Role?," page 27).
"Browser-based tools let us standardize the support of the whole infrastructure
globally," explains Goldman's Uygur. "I can take the management solutions we
develop and just send them over to the 24-by-7 operations staff in Tokyo."
Function Matters
The ability to create an IT department that is centralized functionally rather than
physically also is important for attracting and retaining increasingly scarce
technical talent.
"Our strategy isn't to tell people they have to move because their job is in a
particular location," explains McLees. "That's especially important when you
grow through acquisition as we have and you pick up people with valuable skills
in new cities."
Then there's the scalability issue. As various applications and services grow, IT
managers understand that it no longer makes economic sense to add servers
and memory in small increments. The physical consolidation of processing
power and storage, on the other hand, can offer greater scalability at lower
overall cost-not only reducing the cost of hardware acquisition, but also reducing
management costs.
"You don't see hundreds of RAID arrays floating all over the place anymore,"
Uygur says. "Instead, we're moving to something more akin to a mainframe
approach to storage, with bigger devices that let you support more growth at a
better price."
What it all boils down to is that IT expertise is both scarce and essential, says
John Lewis, First National Bank of Maryland's vice president of data center
operations.
"The so-called independent business units bit off a bit more than they could
chew, because the solutions they bought weren't as easy to support as they
were told by whoever was marketing them," Lewis says.
"Once they realize they have to plug into network infrastructure, monitor
problems, receive data from outside sources and deal with security issues, they
come running back to IT," he adds.
New Rules
Recentralization is affecting how IT does business in several ways. One of the
most prominent effects is the increasing interest in service level agreements. As
IT takes responsibility for infrastructure support away from business units, it's
finding that it needs to improve communications with the new customers.
"One way you can overcome potential objections to the shift in responsibilities
back to a central organization is by making yourself more accountable to the
organizations for whom you are providing support," according to Uygur.
"This allows systems administrators in the business units to become more
confident and comfortable about relinquishing those responsibilities," he adds.
It's also important to understand that centralizing infrastructure and
infrastructure management does not mean a return to the old IT priesthood. In
most instances, business units still want to retain data ownership and take
primary responsibility for how information is used in ways that they never could
in the mainframe era.
"The business units are becoming experts in how they use information at the
same time as they use IT to manage their technology for them," says Younis.
"They're controlling the content and letting someone else take care of the
plumbing."
That's why Younis sees even the concept of the CIO beginning to disappear.
"You don't really need a chief information officer if business units have access
to all the information assets they need,'" he says. "So what you'll have instead is
someone who is in charge of information infrastructure and policy, not the
information itself."
Another major difference between today's centralized architectures and those
of the past is choice.
"With the mainframe, you had to look to a single vendor and a single vendor's
platform for your business solutions,"' says First National Bank of Maryland's
Lewis. "Now, we've created an open marketplace where you have some
negotiating leverage and can get some cost efficiencies."
The trick, therefore, with the new centralization is not to throw out the "baby" of
business unit empowerment with the "bathwater" of fragmented infrastructure.
"We definitely don't want to rebuild the glass house, because users need the
flexibility to work with data and change things to meet their business needs
without having to depend on someone in the data center," says Lilly's Tunstall.
"But we also need to know that our business data is accurate, consistent and
available to anybody who needs it."
That flexibility-and the freedom it represents--is the twist behind the new
centralization.
Lenny Liebmann is a Highlands, N.J.-based consultant and writer who
specializes in technology management issues. He can be reached at
ll@exit109.com.
THE NEW CENTRALIZATION
Here are some tasks and technologies IT is centralizing:
Components: Fewer, larger servers and storage devices save money. Physical
centralization lets IT keep critical servers in more suitable environments.
Planning: Standards can reduce support costs, help ensure interoperability and
cut purchasing costs
Management operations: Staff specialization can promote efficiency, freeing up
technicians for more strategic tasks
Application architecture: By pulling code off PCs and onto servers, IT can aid
support and reduce software distribution/inventory costs. Applications can be
updated quickly, without having to wait for massive desktop rollouts.
Copyright ® 1999 CMP Media Inc.
|
