2006-04-29

Microsoft is no longer a growth investment

In the late '90s, I shifted most of my investments from moderate to high growth into low to moderate growth / moderate income. That's how I was able to weather much of the capital loss unlike many others (I only saw a 5% drop in my worst year), when the .COM bust came and popular growth funds with overvalued stock, like Microsoft, came tumbling down. Although it was clear that Microsoft was making significant investments to stablize its income, most investors who stuck around, or invested in Microsoft after the drop, were okay since Microsoft still had its extreme profit margins on Office and Windows.

Well, that time is now over.

Although some investors are optimistic, Microsoft itself is quickly becoming a software services and investment firm, and no longer a volume software developer. Microsoft has seen losses in its Office and Windows products renewals, gimmicks like software assurance alienate even its most loyal customers as they pay more than if they would just have upgraded incrementally each version and a growing trend of disgruntled Microsoft Gold Partners and certified professionals due to being blamed for patch management issues that are at the fault of Microsoft itself (e.g., "keeping current" caused enterprises to be suseptible to SQL Slammer, because newer patches uninstalled an older fix that would have prevented it).

Without the extreme profits on Office and Windows, which have virtually no cost to reproduced with the same, fixed development costs, Microsoft will never be as profitable as it was just 5 years ago -- let alone nothing like 5 years before that. At the same time, Microsoft is forced to pay recurring infrastructure and, more costly, consulting and personnel costs to provide services to compete with Google and others. So relying more and more on outside investments and more moderate profits on services mean Microsoft's days as a growth fund are over. Microsoft is becoming a regular Blue Chip although their history on income is far from historical Blue Chip. Microsoft's past is filled with investement interest on pure growth only, with 0 dividen payments until recently (just after a very enlightening set of investigations on how Microsoft is not good for our economy by Ralph Nader, among others). It remains to be seen if dividen payments can turn Microsoft into a solid income stock, or if it's really just a redistribution of their existing cash reserves.

Which is where the real trouble begins ...

The only entities allied with Microsoft that are still happy are the Tier-1 OEMs, superstores and the hardware vendors whose products the superstores sell. Unfortunately, as Microsoft's margins decrease and they dip into cash to pay income, that undercuts Microsoft's #1 means of distribution control -- R&D pay-offs. As Intel is finding out first-hand itself -- it's a double-whammy. It's more than just the competition forcing you to lower prices, while you're alienating your consumer who now buys less at the same time, causing your profits to sink. It's the fact that your past, extreme profits have allowed you to spend hundreds of millions in R&D for your partners for their loyalty -- cash you no longer have to do such (without dipping into your cash reserves).

The result is that you lose your distribution loyalty base, and that is the mainstay of Microsoft, just like Intel. The next step is litigation and then counter-litigation. This is Intel today, who has lost the retail distribution with only a 15% marketshare (AMD has nearly 85%), and is losing heavily on the core OEM -- especially servers where AMD jumped from only 5% to 14% in a year where Intel is 33% slower and 50% more costly (let alone AMD has a more compatible, reliable and stable design -- something Microsoft is also fighting Open Source Software, OSS, from a long-term "risk to data" standpoint). Now it took Intel 5 years to reach that point, but don't be surprised in 2011-2012 if Microsoft is starting to feel it just the same.

Because once you stop sending the R&D money, even the unthinkable happens. Like Dell moving away from all Intel-bsaed system components and even considering AMD -- just because its customers want it. And when you try to limit what your partners can do -- like Intel has been with HP (forcing them to sell less than 10% of their servers with AMD) -- you get lawsuits for preventing consumers from buying what they want (because HP has been selling over 30%, and over 50% of HP's customers now prefer AMD for servers). Microsoft is going to start feeling the same from Apple, among others, on the distribution channel -- especially as Apple is replacing Dell as Intel's #1 partner, or the superstores/tier-1 OEMs try to avoid anything but Microsoft.

Now Microsoft, like Intel, will still be around -- long-term Blue Chip companies, just not so much for their past products volume. Microsoft is slowly losing its profits which, again, will directly impact its ability to control its distribution channel -- namely tier-1 OEMs and core providers. But it isn't because someone is trying to directly compete with Microsoft in that distribution channel of their core products. It's because someone was chipping away at their related software services when Microsoft thought it wasn't worth their bother (and outsourced much of it) -- namely Google. But because of Google, Microsoft is finding it's got the double-whammy of affecting money for its distribution channel -- as you'll be seeing over the next 5 years. Microsoft is being forced to compete at margins it's not used to, to fend off Google -- as well as other markets it has (almost without thought) entered like entertainment that has far less profit.

Intel has the same problem, only 5 years newer. It wasn't because AMD was directly challenging Intel at the Tier-1 OEM. They came in through the retail channel. Now they've started socking it to them on the high profit margin server channel -- so much so that even Dell can't ignore the signficiant superiority of AMD's server platform versus Intel. Intel won't go away, but they completely lost the retail market, and are still 2+ years away from partially addressing the server market -- and still not with a solution that is on-par with AMD, who is not standing still.

Intel is a solid Blue Chip, bit not only is it no longer the growth it once was -- it's been negative at times. Microsoft is entering this same period where it's going to see its market cap drop again, as more investors realize that software services are never going to deliver the profits that just software volume does. So like Intel, you will wish you sold your stock sooner, to avoid more market cap loss -- even if you want to hold Microsoft in the future. Buy it back in 5 years, when its cheaper and the income is more proportional to its market cap.

Microsoft is still a sound, long-term investment, just not at its current market capitalization. And that's the problem for people who will hold on to the stock over the next 3-5 years.

2006-04-18

Red Hat-JBoss, Great For Open Source ... But IBM?

In the narrow-minded, over-analyzing of the IT world, you will get mixed reviewed on the Red Hat acquisition of JBoss. In reality, there are at least 2 perspectives.

Great For Open Source!

Absolutely no significant commercial entity even loosely associated with Linux has the history Red Hat does. Everything they do they release as GPL to a fault, with rare exceptions. This even includes allowing their competitors to use and ship the same software, as Novell will continue to do in the case of JBoss.

Sun Microsystems comes closest in GPL donations, StarDivision's StarOffice (now OpenOffice.org aka OOo) and its upkeep (Sun employs half of the OOo core team) being $250M bulk of that contribution. IBM and many use Linux for little more than a product avenue, with extremely limited GPL donations (largely kernel), and sub-$50M donations in open source. E.g., IBM's $1B investment in Linux has been almost entirely non-open source, and even the more recent $100M investment in Notes on Linux is completely proprietary.

Oracle should be considered the Microsoft of the Linux world. In fact, their talk about possibly acquiring Novell does scare me a bit -- especially as Novell has turned SuSE into far more of a pro-GPL, community-focused project than it ever was prior to the acquition.

Even the Apache-Geronimo team should be excited. There's now more code to share with the team under Red Hat's typical approach to copyright and licensing that was not possible as JBoss.

Not the First Time Red Hat Changed Direction

Many are questioning why Red Hat would buy JBoss, when they had already supported Jonas. Red Hat has always wanted to move into the web application space, but didn't know how to proceed. As is typical with Red Hat, they pay a lot of key developer salaries and put them on core open source projects, in the hope they further the projects along. After a year or two, they then start building Rawhide (aka Fedora Development) packages and eventually they make it into their distributions.

Jonas was that direction, until now. The question is why? Well, we'll revisit that in a second.

In 2003, Red Hat announced it was going to build an open source network authentication and directory services stack. This was largely to compete with Novell's eDirectory moves as it released Novell Open Enterprise Server (OES), largely a set of services for SuSE Linux Enterprise Server (SLES), and laid down its plans for a Linux native Netware 7. The core included supporting improvming Kerberos and OpenLDAP developments to add necessary features. Prior to it's Novell acquisition, OpenLDAP was already integrated into SuSE Linux Enterprise Server (SLES).

By fall of 2004, Red Hat completely changed direction. It was already reselling Netscape [Certificate and] Directory Server, the staple LDAP server for enterprises since the late '90s. Sun One uses Netscape Directory Server (NsDS) as its LDAP server, and it includes sychronization support for many other LDAP systems, including Microsoft ActiveDirectory Server (ADS). Despite their work on OpenLDAP, Red Hat bought NsDS (including the Certificate Server) outright, because it was much farther along than OpenLDAP in many areas. As of April 2005, Red Hat secured the rights from AOL to make it GPL open source (sans a few items).

Since then, Red Hat has been able to work on more useful capabilities in Fedora Directory Server (fka Netscape Directory Server), which will (continue) to be commercially known as Red Hat Directory Server (the original rebranding). E.g., one of the nicest additions was the release of a mod_nss module for authentication and complete Apache-Mozilla X.509/PKCS integration -- something severely lacking in the Apache-OpenLDAP space. Things like multi-master replication and ADS synchronization were already part of FDS, saving on the cost of adding them to OpenLDAP.

At the same time, the OpenLDAP team is free to integrate any and all FDS developments, while gaining some "for free."

But Is It Smart For Business?

But Red Hat is still a business, and that has mean wondering why and if they made the right decision for 2 reasons:
1. The timing and costs
2. The possible market fall-out

Unlike the $20M-something "chump change" Red Hat paid for NsDS, JBoss was a bit more costly at $330-420M. In fact, it now goes down as the largest acquition by and for open source beyond Sun's purchase and LGPL'ing of StarDivision and its StarOffice suite (rumored to be around $250M). Why would Red Hat plunk down that much money instead of continuing to develop Jonas?

Was it time-to-market? Would the time it would take to take Jonas to the fore-front be too costly in lost sales? Would that make up for the time-to-market cost, which would be reduced? Or did Red Hat see Jonas not gaining on JBoss and finally decided it was best to acquire them and GPL their codebase?

Or was it because of possible Oracle's interest? I mean, if Red Hat is the GPL flagship, Oracle is most definitely the Microsoft-tactic flagship in the Linux world. If Oracle acquired JBoss, would it not mean that they'd have a strangle-hold over a lot of Linux -- let alone freeze out Red Hat's desire to enter the application space. As much as I love Red Hat's GPL dedication and focus on the community, sometimes they have a lot to learn about VARs and vendor relationships.

Which brings me to the market fall-out to consider.

IBM makes a lot of money off of Red Hat in serivces, especially in the web application space where it sells both licenses of Web Sphere and consulting services. Now Red Hat wants to get into this space, at least in product and integration. That hits IBM in the pocket-book, at least on Web Sphere licenses, even if IBM is still involves in the services. Although Red Hat knows how to approach Linux from the GPL aspect, again, it has shown it does not understand the VAR space. How do they convey to IBM they are not a threat?

It's really not a question of what Red Hat does, but what IBM will ...

If IBM Fears Lost Sales, Red Hat Is Toast

Caldera died 20 days after purchasing SCO. Caldera wanted a split UNIX-Linux model, one with a high-end 64-bit UNIX to off-set its developing, commodity Linux product sales. That directly hurt IBM's AIX/Power sales and its split UNIX-Linux model. So 20 days after their purchase of SCO, IBM undercut them by withdrawing from Project Moneterey with SCO and withholding all source code -- rendering SCO's, now Caldera's, 64-bit UNIX future entirely null and void. IBM didn't need to do this to survive, it would have only cost them a few hundred million dollars in sales -- maybe not even 1% overall. But that's all it took, which rendered SCO a "dead weight" in Caldera's acquisition.

Now Caldera tried to get Linux profitable over 30 months and failed, whereas just a few high-end SCO UNIX 64-bit sales for Itanium would have gave them the cash influx they needed. The rest is now another story, SCO v. IBM. Ironically, IBM hurt competition in the Linux world, yet SCO got stuck with the check because 90% of the Linux community doesn't know how contract law works. Let alone 90% of the Linux community can't tell the difference between a "Non-Compete" clause in a contract and "IP Infringement," which SCO an avenue to expand the lawsuit when IBM didn't settle (like SCO had originally hoped). Now we just have rhetoric atop of rhetoric. All while SCO's case on Project Monetery -- items #50-55 in the original March 2003 filing -- still has merit and SCO is likely to win on some counts against IBM in its planned Utah jury case.

In any case, a very good and pro-GPL Linux company is now dead. All because of not any greediness of SCO (who just needed cash to make it through litigation in May -- an rhetoric avenue created by 90% of the Linux community who didn't realize was little more than a contract dispute in the original March filing), but the original greediness of IBM. IBM, a very large company with the lawyers and liquid capital to starve off smaller companies while they cross them in contracts. A company people blindly assume is "good for Linux" because they have spent $1.1B on maturing their product line, with far, far less actually put into open source -- and almost nothing in GPL donations. Far less than even Caldera-SCO!

But just like the old "Quiz Shows," people don't tune in for the truth, they only tune in for the money.

So far, I haven't seen any reaction out of IBM on the Red Hat-JBoss anouncement. Then again, IBM was pretty silent those first 20 days after Caldera bought SCO. If IBM feels Red Hat is a threat to their bottom line, which this may very well be, you might be reading some seemingly insignificant news about IBM's arrangement iwth Red Hat in the next month. After all, most people thought IBM breaking the Project Moneterey contract with Caldera-SCO wasn't signficant either.

I mean, it was only everything to the future of Caldera.
At $350-420M, JBoss is very much everything to Red Hat, a major investment and strategic move.

So the question becomes, how much do IBM's contracts with Red Hat affect its bottom line? Can IBM use them to adversely affect and/or control Red Hat, or just undercut them completely? I mean, I don't see how IBM couldn't see this as a threat. Will we be scorning Red Hat in 3 years? Will people talk about Red Hat "going bad" and blaming it on everything, and not IBM, like they did SCO?

I hope not. I really hope not. I really don't want to see this again.

2006-04-17

LPI Report on LinuxWorld Boston

From the Linux Professional Institute (LPI) Press Release on
LinuxWorld Boston ...

"LPI at LinuxWorld Boston, April 3-6, 2006 ...

LPI's delegation was led by Jim Lacey, President and CEO and includedthe following LPI staff: Larry McArthur, Area Operations Manager for North America and Asia Pacific; James Stanger, Chairman of the Strategic Advisory Council; Matt Rice, Director of Product Development; Glenn McKnight, Director of Operations; and Scott Lamberton, Director of Communications. Board member Gen Narui also attended the conference and brought an eleven person contingent of LPI-Japan staff, board members and LPIC-2 alumni (see below) ...

Advisory Council meetings:

The semi-annual LPI advisory council meeting was held at the conference and for the first time was split into two meetings: a Strategic Advisory Council (SAC) for discussions on business initiatives and a Technical Advisory Council (TAC) for discussions on product development and certification exams.

In addition to the LPI staff and Board members listed above the following people participated in the Strategic Advisory Council meeting on Wednesday, April 5th at 8:00
a.m.:


* Brian Osborn, Linux Magazine
* Debra Williams-Cauley, Prentice-Hall
* Andy Oram, O'Reilly
* Don Marti, SpikeSource
* Ross Brunson, Novell
* Bryan Smith, IEEE
* Ian Shields, IBM
* Faber Fedor, NJ LUG
* Kevin Dankwardt, IFOSS
* Tobin McGinnis, IFOSS
* Toshiji Tanaka, LPI-Japan
* Shinji Takazawa, SGI-Japan
* Chieko Takahashi NEC and LPI-Japan Board member
* Dr. Haruhisa Ishida, LPI-Japan Board member

The meeting featured presentations from Jim Lacey on LPI's future direction and from Larry McArthur on LPI's "Regional Enablement Initiative". Discussions following
the meeting focused on possible initiatives with LPI alumni.


The Technical Advisory Council meeting was held on Thursday morning and in addition to LPI staff and board members listed above included the following participants:

* Ross Brunson, Novell
* Ian Shields, IBM
* David Allen, Computer Resources Consulting
* Bryan Smith, IEEE
* Jim Nasby, Pervasive/MySQL
* Faber Fedor, NJ LUG
* Kevin Dankwardt, IFOSS
* Tobin McGinnis, IFOSS
* Ricardo Strusberg, ISEIT
* Clark W. Alexander, New England Institute of Technology
* Toshiji Tanaka, LPI-Japan
* Shinji Takazawa, SGI-Japan
* Chieko Takahashi NEC and LPI-Japan Board member
* Dr. Haruhisa Ishida, LPI-Japan Board member

The meeting largely focused on discussions initiated by Matthew Rice, Director of Product Development on plans around the upcoming development of a Level 3 certification."

2006-04-15

The Business of LinuxWorld ...

As I previously mentioned in my first blog entry, Server Hardware at LinuxWorld, it is rare I make these events. My main purpose for attending was the semi-annual Advisory Council and inaugural Technical Advisory Council with the Linux Professional Institute (LPI).

Boston ... Looks Nice ... But Damn It's Cold!

The rumor is that the eastern US LinuxWorld Expo moved to Boston in praise of the state's moves towards open standards over the past year. I think New York would have continued to be a better avenue, and the attendence clearly reflected that. People are friendly in New England, you just have to be prompt and direct in your questions and conversations and not waste their time.

It's no New York, although it's still priced just as bad downtown ...


Since this was out of my own pocket, I didn't feel like shouving out $250+/night to stay downtown. I stayed well out in Waltham, MA, and took the 70A bus route to the MTBA Red+Silver lines. Yes, it was a good 90+ minute route each way every day, but it also meant I got my roundtrip flight + 3 night hotel stay for less than just 2 nights downtown.
I didn't want to lug my executive coat, so I'd normal opt for my Corvette Racing breaker. Unfortunately, I forgot to bring it, and those bus stops quickly became 30+ minutes of weathering sub-40F cold at 6:30am in the morning without any coat.

At least wearing business attire kept me warmer than normal.

The Boston Convention and Expo Center (BCEC) might be dwarfed by the Orange County Convention Center, but it was very easily accessible from the Silver Line (unlike Orlando's sprawling mess) ...


The Floor (Yes, This LWE Was Pretty Dead) ...

The Expo's Floor wasn't heavy with activity, and even far less on the last day, Thursday (when I snapped these pictures), than even on Wednesday. But there were still enough people around to keep the vendors busy enough. I had to avoid some booths because they were way too much marketing coughUnisyscough, but O'Reilly, Red Hat and a few others kept the presentations going with more than just marketing. Of course it seemed no amount of marketing or technobabble by Red Hat could keep everyone from joking that Red Hat's cowboy attire promoting their upcoming Nashville event was actually the Brokeback Mountain theme.

Red Hat and AMD were front'n center. Despite Intel's sponsoring resulting in their name on the badge neckless, the green Tyan-AMD bags were far more popular than the Blue Tyan-Intel ones.



Who I'm Here For ... LPI Day 1 ... Business

The not-for-profit Linux Professional Institute (LPI) has been cleaning some house as of late, and they have reorganized. Gone are the days of just anyone being a sponsor, and exam development now has a formal and full-time lead. Unlike just about any other professional organization in the technology industry that focuses on credentials and certification, LPI does not make a dime on training, which is where the money making pyramid is for most certification-focused organizations. That's why LPI holds the greatest level of international integrity and recognition -- not only in the Linux industry, but for the entire IT industry.

Dr. Strager of Prosoft Learning Corporation, the Master LPI Affliate for North America, mans the LPI booth on Wednesday


The Wednesday morning LPI session was the semi-annual Advisory Committee (AC). Previously, this was a combined business and technical committee. But starting at this LWE, it was purely business. Japan continues to be the leading region for LPI credentials and revenue -- with 55% of LPI's total revenue (more everyone else combined). Germany and then the US fill out the top 3. The concerns and interests of how LPI is addressing the needs of Japan, as well as how LPI can further infiltrate the US and other, emerging markets for Linux, like Brazil, were also discussed.

I stayed largely quiet the first 45 minutes, until alumni and user group involvement was discussed. After hearing a good half dozen viewpoints, I chimed in and introduced myself. I introduced the fact that Linux user groups are very different than any other technical user group -- they are support lists and technically-focused, and not so much social and people networking lists (even programming lists, like Java, are typically social). This also produces a backlash against the certification industry, which causes most people to dismiss LPI as just another "training as a profit scheme."

My initial suggestion was to start leveraging the alumni as "ambassadors" to their local user groups. This wasn't an original idea, but something I plainly stated I stole from the Fedora [Project] Ambassadors effort that uses the few Fedora users to ensure each and every user group is aware of Fedora users in their local group. This offers indirect advocacy in seeing that Fedora users are helped by other Fedora users. LPI's indirect avenue to interest from user groups is to show how the involvement of such LPI Ambassadors can boost attendance and, eventually, membership. If the surrounding industry and their professionals are aware the local user group has certified individuals who can help introduce them to the LPI programs, possibly with some free training, they would be interested in attending. That additional attendance should then translate into "more membership," as I've never seen a knowledgeable IT professional deny him or herself any avenue to "flex their geekdom." Everyone wins.

LPI has largely ignored its alumni and this is a great avenue, with virtually no cost or effort, to not only re-involve them, but market themselves as well. As Ross Brunson of Novell pointed out, Red Hat really does ignore its alumni far worse -- even corporate partners -- and that's something we could address, possibly via the same program. Despite my common tendency to deploy Red Hat solutions, even I couldn't resist disclosing the fact that after I plunked down $749 of my own money to just take the RH302 exam-only, Red Hat didn't even send me a hardcopy of my certificate. They sent me a PDF -- yeah, that really does say something (very wrong)! Especially when a non-profit can even manage that, at 1/5th the 2-part exam cost!

Who I'm Here For ... LPI Day 2 ... Technical

The final day of LWE saw the inaugural LPI Technical Advisory Committee (TAC), with Matt Rice taking over the lead role of Product (formerly known as Exam) Development. Matt Rice and Dr. Stanger were the two individuals who wanted me to make the sessions. We had many off-line conversations about both of our previous involvements (we were both involved with writing LPI exams in years prior).

Dr. Stanger kicks off the inaugural TAC on Thursday in slightly less formal attire


The primary product development in 2006 is the development of LPI Certified Level 3 (LPIC-3). This is very much intertwined with both industry and federal drive in Japan, the region that is clearly leading LPI's proliferation (and revenue). Specialization was regularly brought up, from development to security, but it was partially dismissed as too specific and too time consuming when the goals are to not only define solid objectives for the level, but have much of the exam development completed by year's end.

My largest fear is that with the existence of already 2 levels of "generic/broad" coverage, it's difficult to get any more advanced and more depth at level 3 without getting more specific in application. My suggestion was to take the objectives the Japanese region feels are most important, categorize them into specialized exams that could go to the depth expected at a level 3, and then select those 2-3 exams as the first 2-3 level 3 exams (301, 302 and, possibly, 303). My 2-3 examples were Enterprise Directory/File Services, Security and possibly Internet Services, with another 4-5 specialized exams planned for the future. That would solve my fear that broad and generic 301 and 302 exams could not reach the depth required, while leaving room for specialization, expansion and additional titles (which could result in additional revenue). I'm not sure I made the point in a way all understood where I was coming from, and fear some thought I was "shooting for the moon" in limited time, but I was really trying to "hit the moon now, while still being able to go to Mars in the future."

SELinux came up which seems to be a hot item in Japan. Skills in applying Mandatory Access Controls (MAC) and other Role-Based Access Controls (RBAC) are needed in both enterprise and defense environments, as I can second the Japanese on. But SELinux could be a single, specialized exam on its own. LWE presenter and security expert David Allen also involved himself in the conversation at that point.

As noted by Matt Rice, I had previously recommended that a level 3 LPI "task driven" Security-specific exam adopt the International Information Systems Security Certification Consortium's (ISC2's) existing security Common Book of Knowledge (CBoK) as its conceptual foundation. In other words, we should take several domains from their CBoK and map them to real-world, task-driven Linux application -- since the ISC2 domains themselves are conceptual and not specific, and that's where LPI would come in. This not only gives LPI an immediate blueprint using domains to create task-driven Linux questions, but the mapping gives it a sense of completeness and industry acceptance (something Microsoft completely botched with their Security specialties, by completely ignoring most of the domains in the security CBoK).

While most people at least know of the generic Certified Information System Security Professional (CISSP) and its ten (10) domains generic, most completely ignore the existence of the System Security Certified Practicioner (SSCP), which uses 7 very "real-world" domains when it comes to system security. That includes auditing, [network] authentication (which complements Enterprise Directory and File Services), cryptography, MAC/RBAC (also complementing the same) and 3 other, real-world Linux task-driven mappable domains.

Additional commentary also arose from several organizations providing Linux training from outside the US and Japan, such from individuals with the emerging Linux leader Brazil. With only a half day and so much input, it was difficult for everyone to come to terms on everything, but I believe Matt was made aware of where many professionals and organizations are coming from in their industries.

I reserved myself to write a whitepaper proposal, which I should complete this weekend, that makes all of my prior points in writing. It will summarize most of the technical and development costs (largely in time and considerations of time-to-market), while leveraging much objective focus and exam developments to existing, common practices. I hope it will address the immediate needs of our LPI leading Japanese region, as well as leave room for both expansion and partnerships. LPI has to move beyond just merely revisioning existing exams and there is great difficulty in developing exams at a level 3 depth without detail and specialization. The most sought after skills -- my suggestions (based on my notes from the meeting) being Enterprise Directory and File, System Security (including Enterprise Authentication and other network services) and possibly a 3rd focus in Internet Services -- should be the first 2-3 exams that will make up the initial LPIC-3 certification. Even if they are not branded as specialized exams, but just 301, 302 and possibly 303 to start.

I gaze at the peers to my right as Matt Rice (not pictured) gets the TAC rolling into discussion ...


Hitting the Dot-Org Pavilion

I finished off my Thursday by discussing many aspects of LPI and industry developments with Ross Brunson. I also visited several of the Dot-Org Pavilion booths where the real technical groundwork is not only laid, but written and implemented, by the geeks who make Linux a reality.

The Fedora Project was showing off what Fedora Core 5's new 3D acceleration support can do. Their Composite OpenGL approach (using existing X servers), which is also preferred by nVidia, which is a bit different than the Novell's Xgl OpenGL (completely OpenGL-based server) approach. Both are attempting to bring the features of Apple's QuartzExtreme (also OpenGL-based, only for its native Cocoa-Aqua) to the Free Desktop world. Microsoft's Windows Graphics Foundation (WGF) 2.0, based on former DirectX 10 developments, has been pushed back from Windows Vista (NT 6.0 "Longhorn" Client), and they are going with a far less complete WGF 1.1 release for the "Avalon" presentation, based on existing DirectX 9.

The Fedora Project (founded and sponsored by Red Hat) and Gentoo Foundation (founded by Daniel Robbins, a gentleman I share publication credits with) are the two projects that influence much of my Linux development, integration and usage ...


More on the development front, the GNOME Project (a leading Linux Graphical User Interface system) and related its related Mono (.NET Development for UNIX, founded by the same founder as GNOME, Miguel deIcaza) are two developments I also track closely (especially as I do more Java/.NET development in my career) ...


2006-04-05

Server Hardware at LinuxWorld

I typically never go to conferences unless they are local to where I am currently (be they in Orlando, my home, or in the city local to my current client or employer). But after some appeals from select people with LPI, I decided to make the last 2 days of the LinuxWorld Expo in Boston for select LPI Advisory Council sessions. I'll post a blog entry specifically on LPI tomorrow.

The Hardware Fiend Strikes!

I was pleasantly surprised to get a good glimpse of some hardware, serious PC server hardware no less. That was an unintended bonus, especially since I loathe many shows that seem to focus on consumer goodies. I always asked first, but was able to snap many pictures of several vendor's products, which I will post shortly. The only vendor who showed reservation was QLogic, so I didn't snap any of theirs.

AMD LGA-1207 (Socket-F)

MSI and Tyan were sporting the new LGA-1207 (Socket-F) Opteron platform with DDR2 in 2-way and 4-way. And while others were probably drooling over the "booth babes" at the front of AMD's spread, I was drooling over AMD's impressive "rack" of the latest server mainboards. A good selection of LGA-1207 2-way, 4-way and modular/blade solutions, along with HP and other tier-1 OEMs displaying their offerings.

It was nice to see both ServerWorks HT2000+HT1000 and nVidia nForce Pro 2200+2050 with AMD8132 mainboards. The sweetest 2-way, commodity solution in my eyes was an Arima with two (2) PCI-X 133MHz, (2) PCIe x4 and (1) PCIe x8 slots.

AMD's beautiful rack! No less than eight (8) Socket-F mainboards ...


Unlike Intel, AMD had its guts exposed everywhere. Take this 1U Rackmount 4-socket Opteron for instance ...


And Tyan's ready for Socket-F -- not just 2-socket, but even 4-socket today!


Commodity Storage Solutions

A few vendors had networking and storage on-site. No SAS (Serial Attached SCSI) vendors seemingly made it, so I stopped by CoRAID's booth to get the marketing hype on ATA over Ethernet (AoE). AoE is oversold against both iSCSI and old, multi-target SCSI-2 and I don't think it'll have much life once SAS takes off. I asked their CEO about what he thought of SAS and the renewed multi-target SCSI-2 solution and he side-stepped the question as if I was talking about single-target SCSI-2.

TekRam's Areca was there, showing off their new SATA RAID products. Areca's ARC-11x0 (PCI-X) and ARC-12x0 (PCIe) Intel X-Scale superscalar ARM-based true hardware RAID solutions have seemingly taken over 3Ware's position in the Linux world -- especially since 3Ware lacks a PCIe offering. AMCC/3Ware wasn't present.

MSI didn't have nearly as much as Tyan, although Areca did a nice job showcasing their true hardware RAID line ...


1 and 10Gbps Ethernet HBAs

There were a few vendors with HBAs for FibreChannel and iSCSI. QLogic's reservation on letting me snap pictures wasn't a loss, as they didn't have their newer 1Gbps products anyway (including no PCIe cards). One of the more interesting vendors (who seemed to be wrongly ignored IMHO) was LeWiz Communications, who had both a full-up 10Gbps Ethernet HBA and, far more commodity, a quad (4) port 1Gbps Ethernet HBA in a low-profile PCIe slot no less.

The LeWiz Talon 3008 (which supports both Linux and Solaris). While people talk about commodity NICs with TCP Off-load Engines (TOE) that are little more than send-only (and not very useful), the LeWiz is offering TOE receive (where it's actually useful) in PCIe, in low-profile and in a quad-port for sub-$1K!

LeWiz had FibreChannel PCIe and a very affordable, low-profile, 4-port 1Gbps Ethernet HBA (TCP receive off-load) PCIe ...


And LeWiz even had a new 10Gbps Ethernet HBA along with other solutions ...