|
Dell, SUNY-Buffalo Announce Linux-Based
HPCC
By Charles King
Dell and the State University of New York (SUNY) at
Buffalo have announced what they described as one of the largest clusters of
Linux servers ever deployed at a U.S. educational institution. The high-performance
computing cluster (HPCC) is comprised of more than 2,000 Dell PowerEdge
servers and data storage using a Dell/EMC SAN. The HPCC will be used by the
Skolnick Group in the Buffalo Center of Excellence in Bioinformatics at the
University for human genome research, bioinformatics, protein structure prediction,
and large-scale computer simulations. The cluster consists of 2,000+ Dell
PowerEdge 2650 and 1650 dual-processor servers running Red Hat Linux.
Platform Computing’s LSF 5 workload management software is used to automate
and manage complex computations across the server nodes. The cluster connects
to a 16 terabyte Dell/EMC SAN that uses Extreme Networks’ BlackDiamond
switches for Gigabit I/O connectivity between the nodes. Dell is providing a
variety of services from custom configuration to installation to assist in
the completion of the cluster.
First off, the new SUNY-Buffalo HPCC is an obvious
win for the vendors involved, and likely is particularly gratifying to Dell’s
financial well-being and prestige, and to EMC as an affirmation of its
decision to partner closely with Dell. We believe Platform Computing is also
likely to gain from the project’s high profile, further validating the
company’s new LSF 5 workload management software for cluster and grid
environments, and that Red Hat’s inclusion could further the company’s
efforts to broaden the market penetration of its Linux solutions. All well
and good for the vendors, but does that mean the cluster will be a slam dunk
success for SUNY-Buffalo? From a purely technical standpoint, it should be.
HPCC is a known quantity, and solutions have been sold, installed, and
supported by vendors with extensive HPCC experience, including most of those
listed in this announcement. If there is a question mark here, it is probably
Dell. Though the cluster utilizes field-tested Dell technologies, we are far
from sanguine about the company’s service/support capabilities. Dell has
marched up the hardware food chain by leveraging supply chain acumen and
brutally low prices into market leadership, but the company has yet to prove
that it can deliver adequate, let alone the world class service and support
required by the enterprise customers Dell so covets. Overall, we expect the
SUNY-Buffalo project will test, for better or for worse, the mettle of Dell’s
greater ambitions.
Beyond its implications for the vendors involved, we
also see SUNY-Buffalo’s project as a further indicator of elemental changes
happening in the high performance computing sector. In the past, HPC
installations were the sole prerogative of Big Iron hardware vendors like
IBM, Compaq, and HP. But while those vendors still maintain a commanding HPC
presence, technological advances in microchip design, high performance
networking, clustering, and grid computing have fundamentally altered the
underpinnings of HPC environments. Now instead of the singular,
mainframe-dominated locales of the past, HPC deployments are just as likely
to be made up of numerous RISC- or Intel-based servers deployed across often
widely-separated geographical locations. While this has changed the way HPC
is used, it has also altered the audience for these solutions. While research
facilities such as the Buffalo Center of Excellence in Bioinformatics will
continue to be prime users of HPC, commercial and industrial deployments such
as the IBM/GM HPC environment announced last week are on the rise. Overall,
we expect the technological evolution of HPC will help to drive these
solutions and the advantages they offer further and further out into the
commercial sector, providing capabilities heretofore unavailable to even the
largest of users.
|
|
HP and SAP Target SMBs in Europe
By Jim Balderston
HP and SAP this week announced an alliance designed
to deliver SAP Business One to SMB customers in the EMEA market through HP’s
reseller channel. The alliance will begin its efforts in Germany followed by
Austria, Switzerland, the UK, and Benelux markets. HP will help SAP identify,
qualify, pre-select, and manage local resellers, and join SAP in creating SAP
Business One Centers in the EMEA markets to enable ongoing training and sales
support. The two companies will jointly market the SAP Business One offerings.
SAP Business One offers a range of business information applications,
including general administration, financial accounting, sales and
distribution, purchasing, warehouse management, and partner management.
In a number of ways, the HP/SAP initiative is not
all that surprising, though in others it qualifies as at least intriguing.
This deal makes sense for SAP, which has had less success in the SMB market
largely because it traditional product set is widely — and correctly — viewed
as being expensive, complex, and difficult to deploy. Unlike the large
enterprises that typically have installed SAP systems, SMBs have far less
tolerance for expensive, unwieldy product suites offering every conceivable
function under the sun, and doing so in a fashion that makes end users prefer
root canal work to actually using (or learning to use) the application at
hand. Enter HP and its VARs. HP’s channel offers one viable, organized
methodology for SAP to reach down market in a manner that could well resonate
— at least initially — with SMBs. But the deal is anything but a slam dunk at
this point. If the companies fail to leverage one another’s considerable
experience in a concerted, persistent effort to reach into the SMB market,
the initiative will likely be remembered as a failure whose whole was far
less than the sum of its parts.
We suspect that SAP sees the initiative as an
opportunity to generate SMB sales despite the company’s historical lack of
traction in this market. Many SMBs use applications that provide much of the
same user-facing functionality as SAP offerings, but without the robustness
and complexity. Vendors like Sage have made a solid business out of serving
SMBs with simplified offerings that offer many of SAP’s core functionalities.
Perhaps SAP has decided the time has come to take back some of that market.
At the same time, SMBs using Sage may see this as a good time to move to SAP
for sake of business efficiency and future expandability. While we believe
opportunities exist for HP and SAP to take advantage of these dynamics, to do
so will take more than press releases and alliances. SAP has a well-earned
history of offering products that in essence offer a very large hammer to
drive SMBs’ relatively small nails. Reversing that perception by offering
affordable, usable SMB alternatives will not happen overnight, and will
require both HP and SAP to create and remain committed to a viable long-term
SMB strategy.
|
|
One Size Does Not Fit All, At Least Not
for Software
By Clay Ryder
Novell has announced a new
software pricing program that establishes two additional classes of end-users
and pricing. The new Business-to-Consumer and Government-to-Citizen user
license pricing models are targeted at organizations such as businesses with
many customers, and government agencies and their constituencies, who want to
make Web services and applications broadly available to communities outside
their organizations. This initiative does not affect the Novell standard user
license, which typically applies to the employees, suppliers, etc. of a
business. The Business-to-Consumer user license price is 25% of the standard
user license, and the Government-to-Citizen user license is 10% of the
standard price. In all cases, licensing charges are based on the number of
individual users of a software service, not the number of computing devices
connected to the network. In an unrelated announcement, Microsoft announced
Microsoft Works Suite 2003, its home productivity suite, which includes the
latest versions of Microsoft Works, Word, Encarta, Money, Picture It! Photo,
and Streets & Trips. In addition, the product includes the new Task
Launcher Home with a calendar and to-do list designed to track family
schedules, and My Project Organizer, which integrates all six applications in
the Works Suite. Works Suite 2003 is available for an estimated retail price
of $109 with a $15 mail-in rebate coupon.
While there has been much hemming and hawing about
software licensing fees as of late, the issue has long been a matter of
debate for two reasons. First, licenses provide income to those issuing them,
and second, license fees implicitly propose a valuation on the use of said
software. From a manufacturing perspective, one could argue that bigger
software should cost more, since after all it takes more resources to create
it. But from a user’s perspective, big software where only a fraction of its
capabilities address a user’s actual needs may represent an underutilized or
overpriced investment. This notion of value pricing is hardly new or unique,
but in the world of software, it has rarely been uniformly practiced.
Consider these two seemingly unrelated announcements
from Novell and Microsoft. Novell clearly recognizes that an occasional user
not directly affiliated with an organization would likely derive less value
from a given set of software than would an employee. By lowering the fee for
this class of user, Novell is aligning its revenue more along the lines of
the value it delivers. In a different vein, Microsoft’s latest version of
Works delivers home productivity software at a price that is substantially
lower than its business-oriented cousin, Office XP. But while Works is
targeted for the home user, its word processor is the same as in Office XP,
and the suite is priced lower than a standalone copy of Word. In addition,
Works offers several similar, but not identical capabilities as the low-end
version of Office. While Works is clearly trying to protect corporate price
points for Office, it does so in a way that is inherently conflicted, since
nothing would stop businesses from buying Works solely to get Word, arguably
the most sought after and broad reaching office productivity application.
Thus, the Microsoft approach is to provide different value to the user based upon
total product capability (vendor value) as opposed to user-derived value.
While this seems consistent with the company’s recent corporate software
pricing initiatives, it contradicts Microsoft’s historic focus on the
individual that is manifest in its catch phrase, “Where do you want to go
today?” As a result, in its attempt to devise and deliver different levels of
product value to its users, Microsoft has created a pricing conflict for one
of its most popular products. Conversely, Novell’s approach, which recognizes
that software is not equally useful to all users, avoids such conflict. Given
the current state of the market, we find it reassuring that a leading
software vendor is looking to align its business interests with those of its
customers. Equally disquieting is the fact that THE leading software vendor
seems determined to further distance itself from its customers’ interests — a
notable gaff that in the short term might provide additional revenue, but at
the likely price of long-term alienation of its customer base.
|
|
Another Piece in the Security Puzzle
By Jim Balderston
IBM has announced it is acquiring Access360, a
privately held maker of identity management software based in Irvine, CA.
Terms of the deal were not released, and it is expected to close in the fourth
quarter of this year. Access360 provides software that allows enterprise to
consolidate identity data which in turn allows for automation of employee,
contractor, business partner, and customer access privileges on an enterprise
network. The Access360 technology will be folded into the Tivoli software
portfolio as part of the IBM Software Group.
This latest addition to the IBM security portfolio
continues a path IBM has been following for quite some time. IBM long ago
recognized that IT security is much more complex than a binary in/out
function and is inextricably bound to each application. To achieve the level
of security that enterprises require, security must be much more granular and
specific than merely walling off parts of the enterprise network. With
heterogeneous environments, merged enterprise networks, multiple layers of
partners, customers, employees, and increasingly fragmented access points,
the need for security technologies that attach directly to individual users
could not be greater. Identity management, when manually applied on a
piecemeal basis to an enterprise environment, creates a quagmire of
repetitive tasks simply to add or delete a user. By automating and
centralizing identity management, users can be easily added or deleted, while
ensuring the correct authority is provided to the enterprise network and the
applications that users need to do their jobs. The need for such fine-grained
identity control will not decrease anytime soon, with new technologies like
Web Services and Grid Computing adding to the range of systems and
applications that an enterprise user can traverse. Identity management is
only one piece of the larger security puzzle, yet IBM’s acquisition of
Access360 indicates that is sticking with its strategy of providing a
comprehensive set of interlocking and complementary security technologies
that promise higher degrees of safety and ease of administration than the
security point products that require hand tuning on a per-system, -application,
or -network basis. Of course, none of this will have a great impact on real
enterprise IT safety until security becomes as common and simple to use as a
browser.
|
|
QLogic Drops InfiniBand Development
By Charles King
During a keynote speech at a Salomon Smith Barney
technology conference in New York City this week, QLogic CEO and Chairman
H.K. Desai announced that the company would shelve its InfiniBand switch
silicon due to uncertainties about the market for InfiniBand solutions. Desai
pointed out that the extra help needed in qualification processes around new
technologies, and said the company’s support for investment protection. In
contrast to its InfiniBand decision, Desai said QLogic remains firmly
committed to the iSCSI market and will increase its investment in that area.
The past six months have been a tough time to be in
the InfiniBand business. After being touted as a technology that offered to
fundamentally change datacenter computing, heavy-hitting InfiniBand supporters
Intel and Microsoft decided to leave the market to other vendors, and were
followed to the exits by players including Banderacom, Mellanox, and now
QLogic. Does the withdrawal of Intel and Microsoft mean InfiniBand is finis,
or are other market forces at work here? First, it should be remembered that
despite the media noise generated around InfiniBand the technology has an
extremely narrow range of impact. Basically, InfiniBand is an I/O-focused
solution that bypasses the limitations of PCI Bus technology to dramatically
speed data movement between servers. But the technology also carries
shortcomings including significant range limitations, making it largely
useless outside of datacenter environments. While Intel and Microsoft may
have abandoned ship, datacenter vendors including IBM, HP, Sun, and Dell
remain publicly committed to developing InfiniBand-based solutions. In a
sense, Intel and Microsoft’s decisions make perfect tactical sense. Despite
protestations to the contrary, Microsoft’s datacenter efforts to date have
largely been confined to smaller installations that might not be
significantly enhanced by InfiniBand. Intel’s decision cuts two ways. First,
the company’s current datacenter efforts largely mirror Microsoft’s, though
that will likely change as the company’s 64-bit Itanium solutions come to the
fore. Additionally, the company’s PCI Express (formerly 3GIO) technology
provides an alternative to InfiniBand over which Intel will maintain greater
developmental control.
On a more practical side, the ongoing soft economy
makes InfiniBand’s narrow usability and glacial development cycles a
particularly tough sell, especially for smaller vendors like QLogic. The fact
is that the current drivers for enterprise technology sales are reducing staff
and increasing efficiency, both of which are only tangentially affected by
InfiniBand’s current capabilities. With money tight and sales tough to find,
small vendors will primarily focus their efforts on low hanging fruit and
hope they find enough to survive. This situation is unlikely to change until
the big iron vendors begin shipping InfiniBand-capable servers and storage
products, at which point we expect to see the smaller players jump back in.
But it will also leave the lion’s share of InfiniBand development and
influence in the hands of large vendors that have the financial wherewithal
to focus on InfiniBand’s long-term strategic potential. We also see one
possibly troubling blip on this somewhat stormy horizon. Intel’s decision to
board the PCI Express is likely fueled by the company’s assumption that it
can gain traction and control of the datacenter network as its 64-bit
products become widely adopted. We remember well what can happen when
competing technologies butt heads (Can you say, “BetaMax vs. VHS,” boys and
girls?) and wonder if we are witnessing the genesis of what could eventually
become a PCI Express/InfiniBand train wreck further down the tracks.
|