thinkASG

Sunsets Are Beautiful – Goodbye Sun

IBM is not always the first to market nor do they always enter the market with the best product, but given time, they find a way to prevail.  Like the proverbial slow and steady tortoise that eventually beat the hare trekking through the forest, IBM whizzed past SUN Microsystems in the UNIX server wars.

A Foreign Culture

When IBM introduced the RISC System / 6000 (RS/6000) in the early 1990’s, Big Blue entered the UNIX world of pony tailed geeks wearing sandals – not common place in the white shirt, ties, brick and mortar of mainframe customers.  Although the POWER (Performance Optimized With Enhanced RISC) processor that fueled the RS/6000 was accepted as a decent effort by the UNIX community, AIX was laughed at and scolded for being “non-UNIX.”  IBM had an uphill battle to overcome if the company wanted to be a serious contender in the UNIX marketplace.

 Why would IBM want to enter into a market that seemed so foreign?  The landscape in the 1990’s was changing for IT computing.  More business applications were moving to UNIX as it was deemed “open” and thought to be less expensive to operate compared with IBM’s AS/400 and mainframe systems.  Programmers graduating from colleges in the 1980s and 1990s had been educated on UNIX systems.  The influencers that eventually grew to become IT decision makers were UNIX literate.  The tide was turning and IBM risked losing their customers’ data center business to the hippie culture that was buying UNIX servers.

 At the time, Sun dominated the UNIX workstation market.  The workstations were initially based on Motorola CPUs and later were based on Sun’s own SPARC (a RISC design) architecture.  Apollo, Sun’s biggest competitior, delivered workstations also based on Motorola CPUs.  Apollo was acquired by Hewlett Packard giving HP a decent percentage slice of the market – but SUN was clearly the top dog.  Solaris (Sun’s UNIX operating system) – whether good or bad – was deemed the standard by which all UNIX systems should be based.  IBM’s presence in the market was almost non-existent.

Hare Krishna

As UNIX systems morphed from desk top workstations into rack mounted servers, RAS became exceedingly important.  IBM and HP made some inroads as data center managers realized the importance of “up time.”  Sun took some lumps when their sexy “mainframe killers” (remember the UE10000) proved to be less reliable than advertised.

 Lack of reliability did not seem to hurt Sun.  IBM touted (and delivered) reliable servers and a very stable UNIX operating system (AIX) but Sun was still the “wonder child” of UNIX systems.  It was as if Solaris was a religion… and it was a cool religion.  Believers in Solaris reinforced that it was the “true UNIX” in cult like ways – bad mouthing innovations in the other UNIX variants.  Even the upstart Linux could not sway Solaris worshippers from believing that their UNIX would not reign supreme forever.

“Party on, Wayne.”  “Party on, Garth.”

Sun sold lots and lots of SPARC pizza box servers into the late 1990’s and early 2000’s.  During the “hey day” of the Internet boom, Sun pizza boxes dominated co-location centers filling rack upon rack with 1U servers.  Data centers were a sea of purple Sun systems.  They were good looking and cool to own.  Reliability wasn’t mandatory for pizza boxes because there was another box standing by to replace a broken one.

 While Sun dominated the UNIX market, IBM was trying to get a foothold – something unique – to differentiate itself from “just another UNIX vendor.”  During the mid 90’s, IBM created a massively parallel server called the RS/6000 SP and delivered SMP (symmetric multi-processor) UNIX systems called the RS/6000 J servers.  Although the RS/6000 SP made inroads in high performance computing and some server consolidation deployments, the SMP servers were considered “me too” yawners.  The high point for both systems was the distributed switch technology that IBM incorporated into both systems.  The SMP servers delivered some of the best SMP scalability in the industry and the SP’s switch outperformed any network technology of the day.

 Delivering unique technology in the SP and the SMP servers was a result of an effort started by Lou Gerstner, IBM’s CEO from 1993 to 2002.  Mr. Gerstner noticed that separate factions of IBM were developing great technology but not sharing with other groups within the company.  Lou got the mainframe developers and the RS/6000 developers to hook up and share technology.  The RS/6000 SP was manufactured in upstate New York – IBM’s mainframe neighborhood.

 Still, it wasn’t cool to own IBM UNIX servers.  Anybody that acquired a RS/6000 was a blue suit conformist.

Dot Bomb

Then, just like campus security’s arrival at a fraternity party, the fun abruptly ended.  The “dot com” bubble burst.  Massive spending on Sun systems halted.  Overextended start ups scrambled to preserve capital by freezing IT purchases.  Acquisitions resulted in excess computer systems.  Failed “dot bombs” unloaded their capital equipment to brokers that purchased and resold Sun servers for a fraction of new server prices.

 To meet the frenzied business growth of the late 1990’s and 2000, Sun built up its infrastructure, employee base, and offices.  The dot-com bubble burst of 2001 left Sun over extended and facing massive revenue losses.

Stay the Course

While SUN was trying to reinvent itself after the “dot bomb” bubble, IBM stayed the course – driving toward delivering superior performance while constantly improving their mainstay posture of delivering industry leading RAS (reliability, availability, serviceability) and usability.

 IBM methodically evolved their UNIX system strategy based upon true Big Blue pillars:

• Servers and software worthy of operating mission critical applications
• Protect customers’ investment of legacy IBM systems
• Deliver features that maximize customer investment
• Maximize use of IBM’s technological girth
• Control it’s own destiny

 The POWER architecture continued to be relevant through the 1990’s and into the 2000’s.  IBM continued to invest research and development money into POWER processor technology.  Commitments from manufactures to embed POWER processor technology in their products boosted acceptance of POWER as a mainstream architecture and helped fund R&D.  POWER processor technology appeared in Apple computers, automobiles, video game consoles (Xbox, PlayStation and Nintendo), and networking components. 

 Due to declining revenues, Sun chose to reduce R&D investment in SPARC. 

 IBM chose to continue designing and manufacturing POWER CPU technology.  IBM maintained control of processor technology and server system infrastructure of the pSeries (the RS/6000 name evolved to pSeries) servers.  POWER processor developers working with server designers were able to incorporate features into the CPUs to maximize RAS and overall system performance.  As a result, POWER is not just CPU technology… it is a system architecture.

 Multiple generations of POWER processors POWER, POWER2, and POWER3 provided binary capability with each previous generation of processor – preserving customer investment and easing migration.  IBM’s processor roadmap met projected delivery target dates (and still does) for each successive POWER implementation.

Game Changer

In 2001, in parallel with the POWER4 announcement, IBM introduced logical partition (LPAR) technology to UNIX systems.  This was a game changer that turned the tide for IBM.  LPARs, which utilize hypervisor technology used on mainframe computers since the 1960s (remember Lou Gerstner got the UNIX guys and mainframe guys to hook up?), enable multiple, separate instances of AIX or Linux to run on a single physical server.  Each LPAR instance is completely isolated while sharing server resources with other LPARs.

 IBM finally delivered the unique technology that separated it’s servers from the rest of the pack and gave it a huge advantage over Sun’s pizza box methodology that dominated the industry for so many years.  Essentially, LPARs enabled customers to consolidate racks of pizza boxes into a single server – reducing costs, management, and infrastructure overhead (chassis, racks, power, cooling…).  Server consolidation was introduced to the UNIX world and IBM was the pioneer.  Within a few years took over as the UNIX server revenue leader – stealing the lead from Sun.  Sun has not recovered from the takeover and has faded from relevance.