Duquesne Group Duquesne Group

English version
Recherche et Analyse

Pour éclairer les décideurs dans le domaine de la Continuité d'Activité et de la Sécurité de l'Information, Duquesne Group propose ses analyses et réflexions issues de ses réalisations concrètes en clientèle. Observation critique du terrain par nos experts, contacts permanents avec les principaux acteurs du domaine et expérience vécue dans nos missions de conseil sont nos principales sources. Contactez-nous pour en savoir +
French version
Research

To support decision makers in the management and optimisation of information systems, Duquesne Group delivers in-depth analyses of information technologies, their implementations and their markets. Our research is based on critical observation of the market by our analysts and their on-going contacts with the vendor community, together with hands-on, practical experience from our consulting work. Contact us to find out more


HP, Oracle and Itanium : is HP paying for strategic mistakes?


This conflict, one of the most serious in the IT industry, is indicative of the choices made by vendors over the last fifteen years and of the market’s evolution which is rewarding some players and punishing others.

The first part of this research note is focused on the strategies of the key vendors, first and foremost HP. The second part will look at the different options available to IT users.

A version in French of this article can be found in the blog of Emmanuel Besluau on this site.
Une version en français de cette article se trouve sur le blog d'Emmanuel Besluau sur ce site.



HP, Oracle and Itanium : is HP paying for strategic mistakes?
Oracle recently announced its intention to progressively phase out support for the Intel Itanium processor on all its software. A first very terse statement was later followed by a detailed end of support timetable by product.

Intel for its part said that it had a clear roadmap for the Itanium processor and that it had no intention of putting it into end-of-life.

HP of course is by far the largest vendor of Itanium based servers, with many customers using Oracle software on those platforms. After a short period of verification, HP formally demanded that support be continued and subsequently filed suit against Oracle in a Superior Court of the State of California.

How did things come to this? Flash back…

The reign of "industry standard" and “three way symbiosis”

Since the late 1990's, the model of « three way symbiosis » - which marked the PC industry - spread to the world of servers. This model is characterised by the fact that no one controls everything and that the roles are distributed among players with convergent interests:

  • The processor is the responsibility of a major specialist, with a minor competitor (Intel and AMD)
  • The operating system is in the same situation (Microsoft Windows Server and Linux)
  • Finally, the box itself, incorporating the previous two, is the responsibility of a player who basically manages the logistics and customer contact to varying degrees (the example of Dell and HP is an illustration).

Following this model, servers with "industry standard" x-86 architecture conquered the market.

For this to happen, two factors are needed:

  • Volume: the quantities at stake should allow everyone to achieve return on investment
  • Interchangeability: no player is locked in by the others: HP may sell Intel and AMD based machines, Microsoft as well, Intel works with HP, Dell, IBM etc..

This three way symbiosis worked for a very important part of server sales (90% in volume) but it represents only about half in value, because it coexists with another model in the high-end.

The single supplier heritage of high-end RISC / Unix

The other model in the world of servers (leaving aside the mainframe) is that of RISC-Unix, a model in which a single player has responsibility for almost everything:

  • IBM's controls both the Power processor and the AIX OS in its pSeries
  • Sun offers SPARC processors and Solaris on its servers
  • HP controlled its processors PA-RISC and the HP-UX OS

This world "RISC-UNIX" is very different from the previous model and is today characterised by:

  • The much lower quantities sold, mainly for renewal
  • Price levels – necessarily - higher
  • A significant issue of on-site upgrade and compatibility with existing infrastructure
  • Much less interchangeability among competing platforms
  • A movement to consolidate on a smaller number of more powerful platforms
  • The presence of the most critical customer applications on these architectures
  • The fact that the less demanding applications can quite easily be migrated to smaller platforms like Linux x-86 , which reduces the installed base.

Vendors face a dilemma: as the installed base contracts, with systems becoming more powerful and customers ever more demanding, their own sales revenues decline … and yet at the same time they need to make major investment in the evolution of their product sets.

Is this the equivalent of squaring the circle? Consider the strategies of the vendors.

1998: HP announces the end of life of its RISC processors

This decision was critical and – with hindsight - crucial. Inheriting high-end RISC processors from multiple platforms, HP decided to entrust its future in processors to Intel with Itanium, a technology which is probably (at least on paper) the best one could imagine.

So HP stopped Alpha and PA-RISC and proposed Intel Itanium instead. The vendor evolved HP-UX towards the new technology platform and progressively moved its installed base.

The problem with this decision is that HP’s bet - thinking that other vendors would follow - did not really pay off. Apart from some second-tier or niche players (Bull, SGI) the adoption of Itanium other than by HP was in fact very modest

It should also be remembered that IBM had a project to port AIX on Itanium (Monterey project) at the beginning of the 2000s. This project was abandoned, which probably dealt a death blow to the construction of a three way symbiosis around Itanium.

Still, it is clear that in 2011 HP’s gamble on this point is lost. There is no big movement around Itanium with the quantities needed to drive the adhesion of a sufficient market ecosystem.

HP now finds itself in the uncomfortable position of being almost the sole customer for a product, relying on a supplier whose enthusiasm to invest is limited by the fact that to do so would be mainly in the interest of a single (albeit very large) customer.

Another difficulty: middleware is king !

In the meantime, a fourth type of player made its way and got into this already complicated game: the middleware, with one of the major players being Oracle.

In 2011 it is absolutely everywhere: on x-86 platforms and on RISC UNIX.

Among other things, middleware is interesting in that it allows (at some cost to be sure) to move relatively easily from one platform to another, in using the same middleware. The use of application packages (an ERP such as SAP for example) amplifies this trend.

For the RISC UNIX world, this characteristic (which reduces the barrier to exit) discourages investment by vendors ...because it weakens their hold on the installed base, which is less captive.

Who are the key players in middleware on RISC-UNIX? The answer is clear: mainly Oracle together with IBM (DB2, WebSphere, ...).

However, Oracle has a clear advantage, because it is everywhere: in the world of x-86 and of RISC-UNIX, Oracle products are much more present than those of IBM, which are limited to its own base.

Two other players with no credible middleware offering in their catalog – HP and Sun - were in a weak position around 2005.

Sun’s position was of course greatly reinforced by selling itself to Oracle. Once again, HP found itself alone.

In short, the strategic choices of HP over fifteen years led it to become dependent on another company for its high end processors and almost defenseless against the defection of its installed base via the middleware.

Sun-Oracle has all the cards needed to attack, while IBM has a solid system of defense in place. Where will all of this take the market?

Positions of the three players

Let’s take a look at the situation of each of the players

IBM :

  • IBM has a solid installed base and can defend it
  • it can expand by encroaching on HP’s territory or by recruiting disappointed Sun-Oracle customers
  • it can invest in order to provide the performance and level of service that the market demands
  • It excels in the management of what is “scarce and expensive” (e.g., mainframe)
  • with time, the handicap of its middleware (i.e., low or no availability on HP and Sun platform) disappears
  • It may look like "the best platform for Oracle or for SAP"

IBM has only to continue to invest and be patient ...


Sun-Oracle :

  • its middleware (DBMS, Weblogic, ...) is everywhere
  • it can decide – at least partially - the outcome for HP - Intel: it does not want “three way symbiosis” in the high-end, considering that it lacks economic justification
  • its own installed base RISC (Sparc) is strong but aging, and it is not yet clear how it can evolve
  • it needs to invest in the development of Sparc processors for the evolution of its installed base, but it does not seem to have understood the high end nor to have reassured its major accounts
  • third party software publishers are afraid ...

For Sun-Oracle the challenge is to invest in high-end processors and reassure its ecosystem of software vendors.

HP:

  • its Itanium UNIX installed base is less extensive than those of the others and is worried by Oracle
  • It depends on the goodwill of Intel for the evolution of its high-end processor
  • its faces a risk of defection of its installed base (to Sun Oracle and IBM)
  • it does not control the middleware
  • It cannot by itself ensure the evolution of its Itanium HP- UX base, because it needs partners who are its competitors


HP is stuck and it has to regain control to get out of the current situation. This would seem to call for some major maneuvers. One could speculate that HP might:

  • regain autonomy on Itanium through a deal with Intel, including Intel’s Itanium specialists?
  • finance Intel to develop Itanium more rapidly … convince Oracle? finance Oracle ? get a court judgment against Oracle ?
  • build migration paths towards HP’ x-86 Xeon platforms ?

This last option (go to a platform x-86 with Oracle) is especially significant because it allows the move out of the world of RISC-UNIX. We should however note that:

  • it is possible for any vendor offering x-86, and on this HP is not alone
  • the movement out has already started and in the easiest cases are already done
  • it is not possible for all cases, especially for configurations in the very high end (16 or more processors, large memory sizes) that would require a re-architecture in depth.

None of this would be easy, neither for HP nor for its customers.



In the second part of this research note, we will look at various evolution strategies for IT users as well as the different decision parameters.



Emmanuel Besluau
Mardi 28 Juin 2011

Home Home    Mail Mail    Print Print    Zoom + Zoom +    Zoom - Zoom -    Share Share


Dans la même rubrique :
< >

Vendredi 15 Mars 2013 - 17:00 Grands projets = grands problèmes ?


Duquesne Research Newsletter