x86: 64-bit & SMP

Computers

The following news story "IBM, HP take different tack as Xeon MP moves to 64-bit" has some interesting quotes: http://www.nwfusion.com/news/2005/0330ibmhpta.html

First: "HP has decided to cease production of its eight-way ProLiant DL740 and DL760 systems...". HP is following Dell's withdraw of the 8-socket server space. Apparently Dell and HP believe that there is little market demand for more than a handful of threads (today an OS schedules one thread per core or hyperthread context). Or, could it be that their Operating Systems of choice (Windows and Linux) simply can't (yet) scale to larger thread counts? Hey Dell & HP... you might want to check out Solaris 10. A million of your prospects have downloaded and registered this OS in just the last two months! And it runs just fine on your (small and large) x86/x64 servers, up to hundreds of threads.

Second:  Andy Lees, corporate vice president with Microsoft's server and tools business, said "If you run a 32-bit application on 64-bit Windows [Windows Server 2003 x64 Edition] on 64-bit hardware, you'll get about a 5% bump in terms of performance," he said. "If you go ahead and add 64-bit [application] capabilities, then things get dramatically better."

Hmmm. This is an interesting admission that 64-bit might actually be worthwhile. It is (not really) amazing that up until Microsoft (x64 Edition) and Intel (EM64T) had decent 64-bit offerings, that they told the world that 32-bit was all that anyone would need for the foreseeable future - except maybe for huge databases and extremely large memory footprint compute jobs. I guess "foreseeable" means until we can field a team. Oh, by the way, Solaris has been 64-bit forever (in Internet years), has unmatchable security features and reliability, and a bundled virtualization technology that alone is worth the price of admission (oh yeah, it's free).

Combine small and high-thread count performance, security, reliability, and virtualization... and Solaris 10 will allow you to stack multiple applications on a single x86/x64 server with confidence. All of a sudden an 8-socket server (with 16 high-performance cores) looks like an important sweet spot for driving utilization rates up and operation cost and complexity down.

HP and Dell have withdrawn from that space (a strategic blunder I believe). It'll be interesting to see who steps up to claim that prize!

March 30, 2005 03:32 PM EST Permalink

Itanic: Davy Jones' Locker

Computers

In the year 2000, just as the first Itanium processor from Intel hit the market, IDC predicted that 2004 Itanium server sales would hit the $28 billion mark! But IDC missed their projection slightly. They were off by about $26.6 billion, or ~95%. Ouch!!

Of the few Itanium-based servers that were actually sold in all of 2004, HP lead the "crawl" and accounted for 76% of them. But HP, as of mid-2004, joined Sun and IBM in the Opteron-based server market, so expect Itanium sales at HP in 2005 to slow at a faster rate than HP's general server sales numbers. IBM came in 2nd with 10% of the Itanium market, but has strongly hinted that they are killing off their Itanium-based server offerings in favor of Opteron, Power, and traditional Intel processors. Dell captured 3rd place with just 5% of the tiny Itanium pie, and so far Dell has resisted selling Opteron-based servers... but how long will Michael watch from the sidelines?

For those who like to look under the hood, it seems to me there are three server-oriented processor families that deserve attention and will still be important in 2010:

1. Sun's (and Fujitsu's) SPARC-based CMT families (US-IV, Olympus, Niagara, Rock, etc)
2. IBM's Power family (Power4, Power5, Power6, etc)
3. AMD/Intel's x86/x64 families:
1. Opteron/AMD64 [Egypt, Italy, etc]
2. IA-32/EM64T [Nocona, Potomac, Smithfield, Tulsa, etc]

It will be fun to watch. They all have well funded R&D, aggressive rates of innovation, compelling roadmaps, and market/ISV traction. I believe all three horses will be in the race five years from now, but only two will be perceived as the market leaders. Unpredictable market dynamics and execution challenges will likely cause one of the three to stumble and fall behind. But anyone's guess as to which will stumble would be just that - a guess. Intel can survive a $25 billion dollar mistake, and learn from it; and AMD is actually delivering new processors faster than their roadmaps suggest (an amazing feat for a processor design shop)! IBM's roadmap and processor technology look great, but massive CMT could explode and their Cell Processor could turn into the next Itanic for server applications. Sun has Olympus to compete with Power6, and very exciting new yearlings (Niagara and Rock) that could, well, Rock the world soon. Single-threaded deep pipeline performance processors, throughput-oriented massive-CMT chips, and price-efficient desktop/presentation CPUs are all up for grabs. I doubt one horse will win the Triple Crown. Stay tuned.

Of course, OS traction will dictate this to some degree (Solaris, Linux, and Windows64 are all interesting candidates), as will J2EE -vs- .NET adoption and COTS app support. I think that security and efficient/reliable virturalization technology will be key drivers of platform selection in future years.

The one thing we can predict with near certainly is that Itanium (aka: Itanic) is headed to Davy Jones' locker.

March 29, 2005 10:33 AM EST Permalink

Life, Death, and the Future

Personal

Datapoint #1
This year, 2005, will mark a pivotal moment in human history - the year that the global crude death rate will start rising for the first time since the dawn of civilization some 10,000 years ago. The crude death rate measures the number of people per 1,000 population who die in a year. That figure has been falling steadily, from about 40 in pre-agricultural societies to around 8.7 today. It will now start to rise because health practices and infant survival has lead to an aging population, thanks to better nutrition, greater access to medical care, improved sanitation, more widespread immunization, better health education, etc. The UN projects the crude death rate will eventually increase to match the crude birth rate at 11.4 in 2075 with a population of 9.22 billion, and then grow beyond the crude birth rate causing a population decrease.

Datapoint #2
As of this morning, there are 295.735M people in the US, about 4.6% of the 6.427B in the World.
Population Clock: http://www.census.gov/main/www/popclock.html

Combining the two, it's amazing to consider that over 150,000 will die today around the world. That is approx how many died in the recent Asian tsunamis (125,598 confirmed dead, 94,574 people still listed as missing). Every day!

Someday, each one of us will die. Hopefully we'll have made a difference in this world. If we're lucky, we might even be remembered  by a few for a couple generations as someone who provided for and loved their family and friends, who cared for and served their communities, who created opportunities for those around them, who lived by strong principles, who did not hesitate to sacrifice their own plans or leisure to help those in need. I was encouraged to hear Jonathan Schwartz, Sun's President and COO, a brilliant and successful business and industry leader, and father, explain that his kids are his #1 passion in life. Here's a 2 minute audio clip of a recent interview.

Still, even the best of us will be utterly forgotten within a short time. Do you know much of anything about your great-great grandfather? His name, what he did, or even where he is buried? Very few of you I suspect. Unless maybe you go home and look it up in a dusty book somewhere.

Terri Schiavo might be one of the ~150,000 that die today. While most deaths are painful at some level, some deaths seem to impact the global consciousness more than others. And maybe alter our thoughts or even our laws to some degree. But, even Terri will be soon forgotten by almost all of us.

There is one death that stands out among all the rest. Today is Good Friday. Probably the most profound and memorable and impactful death in the history of mankind occurred about 2000 years ago. Even our modern calendar is based on this man's life and death. Our currency contains the year since he lived, and a motto attesting to our national devotion. Jesus was nailed on a Roman cross to die slowly and painfully to appease the public will and the will of God, and even the desire of Jesus. Why was this "Good"? Certainly not because he deserved to die. Certainly not because it was a pleasant or dignified death. We call this Good Friday because of the saving effect his life and death has on us and our hope for tomorrow and beyond. Our great-great-great grandkids will be celebrating his death (Good Friday) and resurrection (Easter) this same time of year, and yet will have forgotten we ever existed. If we truly care about our family, and our future generations, we can best serve them by demonstrating our faith in action.

This is a related true story that underscores the importance of setting proper priorities and passions before it is too late...

Pete Maravich is regarded as one of the top five basketball players of all time. After his retirement from basketball, Pete found true happiness in Christ. "He gave his heart to Jesus Christ, and for the next five years, he was on fire for the Lord." Learning of his conversion, Dobson invited him to appear on his radio program in 1988. After the interview, they played a game of pick-up basketball with several others. When the game ended, Dobson turned to Maravich and said, "Pete, you can't give up basketball. This game means too much to you." Maravich explained he had experienced pain in his right shoulder for more than a year, but now it had disappeared. "I feel just great," he said.

Those were his last words. Maravich collapsed on the basketball court, and minutes later, died in Dobson's arms. "Later that day, I sat down with my son Ryan, who was 17 at the time," Dobson said. "I told him that what happened to Pete wasn't an isolated event. This is the human condition. This is all of us. It will happen to me some day. Pete Maravich didn't have an opportunity to speak with his family one last time. But I want to tell you, be there. On resurrection morning, be there. I will be looking for you then. Nothing else matters. Be there."

Two years later, Dobson suffered a heart attack on the same basketball court.

March 25, 2005 02:32 PM EST Permalink

The Rat Race: Treadmills

Exercise

I've really come to enjoy running on the StarTrac Pro Elite treadmills at LA Fitness. Their "Impact Absorption System" does seem to reduce shock to my knees and back (compared to running on the concrete golf cart paths in my neighborhood). Having a predictable climate (it doesn't get too hot, cold, wet, windy, etc) and extended hours of operation (5am until midnight) helps remove possible excuses. The treadmills even have built-in fans to keep me cool, and there are plenty of distractions to pass the time (TVs tuned to FM frequencies, a view of the racquetball courts, etc).

Over time I've settled on three different running programs to increase the variety and keep running challenging and interesting. As a side-effect, more serious runners tell me that these three types of running compliment each other in terms of achieving peak performance...

Heart Rate is a decent indicator of health and intensity. I've found three different HR metrics are useful to track. I wear a Polar HR Monitor chest strap while running - which the treadmill detects and displays.

1. Resting HR - just as you wake up in the morning [I'm about 47 bpm]
   
2. Peak HR - at the end of your run (or during the last intense interval) [I'm about 160 bpm]
   
3. Recovery HR - during the 2 minutes immediately following intense exercise (while walking @ 2.5 mph) [I'm about -50 bpm]
   

Here are my three treadmill programs:

1. 5K Road Race
This one is easy to setup... Just set the speed (eg: 6.8mph) and incline (I keep it level) and go. Run for 3.12 miles, followed by a 2.5 mph recovery walk.

2. Dynamic HR Intervals
This program automatically adapts to real-time bio-feedback... I tell the treadmill I'm 20 years old so that it'll let me use more aggressive settings. I cycle between 142 and 160 bpm (HR), with a maximum speed of 7 mph, for 30 minutes. I'll increase the intensity and duration over time. The treadmill detects my HR from the chest strap and slowly ramps up to max seepd (currently 7 mph) and then increases the incline (to about 5% for me, at this speed) to get my HR up to 160 bpm. Then it slowly ramps down to 0% incline and a slower speed until I recover to 142 bpm. This repeats several times. The time seems to go by more quickly than running at a set pace.

3. Sprint Intervals
By using the CUSTOM interval program, I can do moderate wind sprints. Not flat out like you might do on a track, but fast enough to get anaerobic . I'm just starting these, so I'm starting easy. This is my current sprint routine, which I'll increase over time. I'll probably alternate adjusting sprint duration and speed, and/or incline. Or I might tweak the recovery phase.

Total Distance: 4.0 miles

• Segment #1: 0.2 miles @ 3.5 mph  (warm up)
• Segment #2: 0.4 miles @ 5.5 mph (warm up)
• 8 Repeats (for i=1 to 8)
  
• Segment #2i+1: 0.2 miles @ 4.5 mph (interval recovery)
• Segment #2i+2: 0.2 miles @ 8.5 mph (sprint)
• Segment #19: 0.2 miles @ 2.5 mph (cooldown)

March 24, 2005 10:28 AM EST Permalink

Good Enough -vs- Gratuitous Upgrades

Computers

Sun offers a really cool thin-client called the SunRay. Check out this flash! We've got 30,000 or so running our desktops throughout Sun. Zero-admin, highly-reliable, energy-efficient clients have saved us millions and driven up productivity. Many of our customers are running these as well. There isn't much to the device... No OS, no disk, no fan, no viruses, no patching, no state... you can almost think of it as a remote/networked frame buffer on steroids. Coupled with USB peripheral support, mobile session capability, Java card security, DoD approved multi-compartment support, VIOP telephony, this is a device that deserves all the attention and acceptance it is getting.

Using Tarantella, Citrix, or other techniques, this device can even display full screen Windows (indistinguishable from a Windoze thin client) if desired, or it can run "Windows in a window" from a native GNOME Linux or Solaris desktop. With the Java Desktop System's integration of hundreds of bundled apps (StarOffice [MS Office], Mr. Project [MS Project], GIMP [Photoshop], Evolution [Outlook], etc, etc) some are looking at the oppty to stop payment to Redmond.

Whatever your choice of display and environment, just pull your Java Card (your session is preserved on the server) and reinsert it later at home, or the next day in another office, and your session will "instantly" pop up in front of you ready to continue your work.

However, a customer recently expressed a concern that the SunRay isn't powered by the latest processor technology, and isn't populated by a huge bank of RAM. Hmmm. I wonder if this person might also consider writing to and asking:

Norelco why their electric razors are powered by two AA batteries! When MegaRaz offers your choice of 220V 3-phase or dual-feed 30A single-phase units that can rip thru facial hair and auto-exfoliate the top layer of skin in record time.

Panasonic why their microwave ovens are still powered by radio-wave emitting magnetrons. Don't they know that MicroRad now offers lead-lined plutonium-powered resonant-coupled chamber ovens that can cut food prep time by a factor of 50 over obsolete microwave ovens?

Kenmore why their refrigerators have not kept up with the times. That DeepFrz and many others now offer a turbo-switch option that circulates liquid hydrogen to drop the freezer compartment temp to near absolute zero, extending food storage times to future generations. Many use this feature to preserve small pets during vacations, eliminating the need for pet sitting or boarding.

Those were designed to be funny, and to make the point that often engineering makes design choices that are "good enough". The SunRay has to have enough power to paint pixels. And it does. Future versions might require more capable processors to handle stronger encryption at faster network speeds, 3D Acceleration, etc. But gratuitously incorporating leading-edge technology into a design can increase cost, heat, power, noise, and instability with no added benefit. Be careful what you ask for... because you'll end up paying for it. Requirements should be linked to the business value they provide,  and not to an emotional "got to have it just because" craving that is fueled by consumer marketing campaigns.

March 23, 2005 08:20 AM EST Permalink

Right to Life

Personal

Mike Duigou got me thinking about the whole "Right to Life" topic. So has CNN. Allow me to post a brief personal commentary on this important subject.

First of all, it seems self-evident to me (and the founders of our nation and our universe) that humans have an inalienable right to life. Webster defines inalienable as "incapable of being alienated, surrendered, or transferred". Here are some foundational texts that support this concept and upon which our national identity and our laws and our ethics are based:

Declaration of Independence
We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty, and the pursuit of Happiness.

Constitution of the United States of America
Amendment V & XIV: No person shall be ... deprived of life, liberty, or property, without due process of law...

The Hippocratic Oath (has provided moral guidance to physicians for 2500 years)
I will not give a fatal draught (read: deadly drug) to anyone if I am asked, nor will I suggest any such thing. Neither will I give a woman means to procure an abortion.

The Holy Bible (King James Version)
Exodus 20:13   Thou shalt not kill. (read: murder)

Unfortunately we have a national history of legally depriving (alienating) entire groups of people of their basic rights based primarily on prejudice (slavery) or convenience (abortion). We've since corrected our moral failure in one of these areas. Tragically, we seem to be on the verge of adding another disgrace to our national record... permitting the taking of life based on a misguided sense of compassion (involuntary euthanasia). Holland decriminalized euthanasia in 1984 and today one in five assisted suicides is without explicit consent, and this form of death has risen to 3% of all deaths in Holland. It's not inconceviable that America might someday  give doctors the right to terminate the life of deformed children and the eldery without consent.

There are also several reasonable and legitimate ways in which our legal system provides for exemptions to an individual's right to life. For example, a person may choose to waive that right (eg: a living will, voluntary assisted suicide) and have their life terminated by another. Or that right may be withdrawn by a court based on a conviction of a capital crime (eg: death sentence). Or that right may be justly and forcefully preempted thru police action, acts of war, Presidential Order (sanctioned assassination of an enemy of the State), or (self) defense. Finally, a guardian may, under certain circumstances, make the choice to terminate the life of someone under their care (eg: a brain dead child or spouse).

I do believe that long-term medical life-support should be only applied to sustain life when there is a clear expectation of a reasonable recovery of a desired quality of life. Yes, I know those are ambiguous terms, wide open to (mis)interpretation. Naturally there will be cases about which reasonable people with no hidden agenda or conflict of interest might disagree. It is in these cases, where there is no living will (or decision by a devoted/informed guardian) and there does not exist compelling evidence of a persistent state of unconsciousness, that a society must choose life over death. It is simply not our place to play God. While it might be technically possible to sustain the biological "life" of a child with no brain wave activity using an iron lung, a kidney dialysis machine, a catheter, and heart bypass pump, no rational/loving person would desire an extended expression of technology overriding nature in a case like this.

However, the present case of Terri Schiavo stretches our ability to divine an ethically appropriate resolution to a life that is clearly demonstrating some level of cognitive ability. If our society prevails to involuntarily euthanize her, we should at least follow the humane process we use to terminate those on death row... inject her with pain killers followed by quick acting drug to stop her heart. How can we allow an innocent invalid who can't speak or scream in pain to starve to death over a period of one to two weeks! Try going without water for a day or two and see how you feel. That's pure torture and is as sick and barbaric as ripping apart a perfectly viable full-term baby from the womb their mommy. Yes, those are emotional words. For an unimaginable act.

Right to Life can be a complex issue at the boundary conditions and corner cases. However, those are extremely rare. The vast majortiy of cases of slavery, abortion, and euthanasia are easily resolved by simply considering the value of life and the motives of those who desire the termination of life or the alienation of rights.

March 21, 2005 09:00 PM EST Permalink

Competitive Takeout Conundrum

Puzzles

Sun Microsystems has kicked off a new sales initiative. To stimulate new revenue, 200 small but lethal "A" teams have been assigned to as many prospect accounts around the world - small but strategic/growth accounts with no Sun kit to date.

The teams were given an initial task to assess these accounts w.r.t. competitive server installed base, to help plan assault tactics. Here are the complete results of that study:

• 104 accounts have IBM, 81 have HP, and 125 have Dell.
• 39 accounts have IBM and HP, 58 have IBM and Dell, and 50 have HP and Dell.
• 16 accounts have ONLY IBM and HP on their floor.

A Sigma Black Belt (an expert in SixSigma methods) was assigned to analyze the data.

How many of the accounts did the Black Belt find that did not have any servers from IBM, Dell or HP?

March 21, 2005 09:22 AM EST Permalink

Sebring & Sun: The Races

Personal

My son, my dad, and I went to the 12-hours of Sebring race yesterday! It was a great time... perfect weather and lots to see and do and experience. Oh, and there was a car race as well :-) The Audi R8 has owned this race for years! The above isn't the actual car that won, but the same model. It beat out Corvette CR-6s, the new Aston Martin DBR9s, Porsche 911s, Maserati MC12s, Lola EX257s, Saleen S7Rs, Dodge Vipers, Ferrari 550 Maranellos, and others,  After 6 years, Audi will debut the R10 next year, which should secure the gold for another generation.

If you haven't been to a car race before (I must be one of the few that hadn't), Sebring is the first and longest race in the annual American Le Mans Series. It's a "street race" (unlike Daytona's NASCAR style) on a 3.7 mile loop on an old WWII airfield with flat hairpin corners that require strategic down shifting and braking. The longest straight segment allows cars to approach 200mph! Average lap speed can exceed 120mph. This year 37 cars started (and 18 finished) the 300+ lap race. Each car has a pit crew and several drivers. There are four classes of cars (each capable of different top and cornering speeds) racing together, making it a very different race than, say, the Indy 500.

Like a golf tournament, there are 4 days of racing (Wed-Sat) leading up to Saturday's main event. Many thousands of people lined up on Wednesday to drive their RVs and U-Haul trucks and School Buses and other interesting forms of transport onto the grounds for an extended campout/party. Some built scaffolding along the course on which they placed sofas! It's March after all - Spring Break time. People watching was as entertaining as the race itself! A friend of mine has gone every year since 1959.

There were also many interesting "sponsor" displays, including race cars in various stages of (de)construction, allowing fascinating views of the internals of the cars.

As we watched the cars fly by our viewing area (we set up our canopy and chairs on a grassy knoll between turns 6 and 7), my dad and I discussed the stresses involved in engines that propel these cars ~1300 miles at an average speed of well over 100 mph, much of it accelerating out of a total of over 4000 turns over 12 solid hours. Each piston cycles about 5 million times! We wondered, at red line speed, how fast those pistons move (12 hours is a long time to talk :-). Are those piston heads traveling faster (inside the cylinder) than the car itself? Reflecting on riding my bike at 20mph, I knew my feet (acting like a piston) don't move that fast... so I guessed that car pistons at red line would move slower than a car's top speed.

Well, I had to figure it out. The math is easy. Obviously piston speed is tied to RPM and Stroke Length, and just indirectly related to the car's forward motion (you can spin the engine in neutral at a standstill). At one of the displays I found out that the typical stroke length for these types of engines is 2.5-3.5 inches, and the typical red line is 8,000-10,000 RPM. Note that RPM measures the crankshaft rotation rate, not the camshaft, which is rotating at half that speed, as shown in this animated GIF:

Since speed is distance/time, and since the piston head travels 2*StrokeLength (up and down) for each cycle, the *average* piston speed is:

Therefore, a car traveling at nearly 200 mph, at the engine's red line, will have pistons traveling at an average of only around 51mph! I was right.

But wait... that piston starts and stops 18,000 times every minute, accelerating to the next stop just 3 inches away (at a huge "G" force). Average speed does not really answer the question. I had to figure out the peak piston speed. Since it literally explodes from a dead stop (it's a combustion engine after all), possibly the peak speed exceeds the speed of the car?

It turns out that the peak speed equation is complex, with sin/cos kinds of rotational acceleration factors. I took the easy way and looked up stats from several types of race car engines using Google. The ratio of peak to average piston speed is consistently very close to 1.6:1.  So at Sebring, the pistons never traveled faster than about 82 mph.

Even a Formula 1 (NASCAR) car, with an engine that can red line at 19,000 RPM, that has a stroke length of about 1.65 inches and a top speed ~240 mph, will have pistons that average: 2*19000*1.65 = 60mph, and peak at: 95mph. See footnote below.

So, why did I include "Sun" in the Subject Title? Check out the Rearview Mirrors. We're a sponsor!

http://lordcaffeine.com/wordpress/index.php?cat=2
At 19,000 rpm, 316.7 revolutions and 1,583.3 ignitions take place each second in the BMW F1 engine. 9,500 engine speed measurements are made, the pistons cover a distance of 25 metres, and 550 litres of air are drawn in. In the P84, maximum piston acceleration was 10,000g. Peak piston speed was 40 metres per second.

March 21, 2005 08:10 AM EST Permalink

IDC / Statistics / Lies

General

I once heard that there are three kinds of lies: Black Lies, Little White Lies, & Sermon Illustrations  :-)

I'd like to add one to the list... "Precision Statistical Estimators".

In a recent edition of Information Week (Jan 31st, 2005, page 22), Paul McDougall quotes an IDC study that suggests that "software-on-demand" sales from full service IT vendors (like IBM, Sun, etc) will grow at a compound annual rate of 40.6% through 2008.

Cool! However, this rant is not about Information Week, or about Paul, or about IT Vendors, or about "software-on-demand". It is about the absurdity of high-precision swags. Projecting a new and speculative market tend to decimal point accuracy? Come on IDC, get real!!

We do need smart analysts to look at all the driving factors and give us reasonable estimates for various market shifts and opportunities. However, it would be much more credible, and useful, if they would say something like:

We project (with 90% confidence) that market demand for XYZ will exhibit a compound annual growth rate of 35-45% through 2008.

Sure, I might just plug 40% into my model. Or, I might have a more sophisticated model that looks at what-if scenarios at both boundary conditions (35% and 45%). But if you tell me that 40.6% is the answer, I'll tell you that you don't understand the question.

Last night on the news, the father of a missing child was asked if he considered asking a psychic to help find his daughter. He said that if that technique really worked, there would never be any missing children and everyone would win the lottery! I appreciate his rationality, and pray for his situation.

But, apparently, IDC thinks they can predict with stunning precision. Hey IDC - got any lottery numbers for me???

March 18, 2005 08:58 AM EST Permalink

Architects & MyersBriggs [INTP]

General

If you have 5 minutes, you can get a reasonable idea about your personality type here:
http://www.haleonline.com/psychtest/index.php

I'm a strong INTP. Each of the four categories have two choices, so there are 2^4 = 16 different profiles or personality types (see graphic below). You'll fall into one of those sixteen and will pretty much be that for the rest of your life - that is who you are. While it can be useful to know your own type, it is probably more valuable to know that of your teammates and companions - and use it to leverage the strengths that each brings to the table.

Since each of the four categories defines a "preference", you'll probably find that you have some tendencies described by both choices. However, if you really think seriously and honestly thru the questions, you'll find one more strongly defines your comfort zone and/or typical mode of operation when given a choice and the freedom do as you please. Note that you might associate negative meaning to the descriptors: "introvert" and "judger". They don't mean shy and critical/mean in this context, so focus on the introspective questions in each category, rather than the one word descriptors.

For me, the E/I category was more of a toss up. But I'm a strong NTP. Scott McNealy and Albert Einstein are/were INTPs too. It turns out that most Architects fall into this profile type as well.. That's good, I guess, because I'm an IT Architect. The role and personality seem to match and I love what I do. But as I suggest below, diversity is healthy.

As an INTP, I tend to listen more than I talk, and then think before talking. I tend to look at the possibilities and big picture, and  then figure out how to get there in interesting and creative ways. I am fact/truth based and am concerned with efficiency and forward progress more so than sensitivity or political correctness. I enjoy coming up with creative solutions and starting the process (architecting) more than I enjoy closure (implementing). And along the way, the "NT" in me enjoys a good debate about the best approaches and the logic that supports the decisions.

However, one of the benefits of knowing your own "type" is to explore possible challenges and how others might help complete the team. An INTP often spends too much time talking/thinking about what can be, and can be light on filling in enough of the details. We tend to over think and analyze.. always considering new options and optimizations, sometimes not executing promptly against the time line as desired by an "SJ" type (eg: a Project Manager). A "P" types get their max adrenaline rush and surge of productivity when the drop dead time is at hand.... whereas the "J" type maps out a step-by-step plan and manages to that (fixed) schedule.

Clearly, there is an advantage to a team that consists of a diversity of types.

I just completed the first 1.5 hours of the Facilitated Mentoring program that initiates a mentor/mentee relationship at Sun. I'm the mentor. My mentee is an ISFJ - almost my polar opposite. This should be a rich/rewarding experience for both of us. I highly recommend it.

March 14, 2005 02:05 PM EST Permalink

SOA: Debating our CTO

Computers

I have the utmost respect for our CTO of Enterprise Web Services, John Crupi. He is a great guy and one of our sharpest arrows. If you get a chance to hear him speak, you will enjoy the time and take away valuable insights. John recently joined the BSC community (blogs.sun.com) and posted a brief intro to SOA. Welcome John! I look forward to future updates on this topic on your blog.

Me, I'm a Lead Architect with a background built on consulting and systems engineering primarily at the IT Infrastructure level, focused on most of the solution stack - up to but not generally into the functional business logic or S/W app design space. Prior to Sun I spent years as a programmer translating business requirements into S/W solutions... but that's been a while.

With that context (the fact that I come to the table with certain biases and experiences that color my perceptions, and I suspect John does as well, to some degree) I'm going to suggest that maybe John is slightly off-base w.r.t. his premise about SOA and IT / Business Unit (BU) alignment. In the spirit of extracting deeper insights and clarifying positions, I'm going to challenge John with an alternate view (a debate), and ask him to either agree with me or defend his position. Hmmm...is this a Career Limiting Move - publicly challenging one of our Chief Technology Officers? No... not at Sun. We encourage our folks to question assumptions and even our leaders, resolve/align, and then move forward in unity. Okay, with that:

John, you suggest that: one of the critical success factors for SOA is a tighter relationship/alignment between Business Units and IT. In fact you say we can not even do SOA without effort on the part of the Business Unit.

Now I could not agree more that Business/IT alignment is absolutely paramount. The lack of business focus and alignment is one of the top reasons why so many IT initiatives fail to deliver or meet expectations or provide a higher return to the business than its cost. I've blogged about that very topic.

However, that alignment, IMHO, is not related to SOA. In fact, I believe there are benefits to isolating service construction techniques from the consumers and owners of those services. To reuse the power utility metaphor:

You don't care how S&L built the power plants that deliver your electric service, or how power distribution provisioning logic taps into multiple grid suppliers and peak-load gas turbines. You simply have specific service level and financial demands, and expect a quality experience when/if you have to interact with the service desk to resolve a dispute, request a change in your service, or report an incident.

There are two primary components to IT... the design/development of services, and the opertaion/delivery of services.

"Business - IT Development" alignment is driven by business requirements (functional, service level, cost, time-to-market, etc). SOA isn't a "requirement", but a technique that helps IT achieve the desires of the business to support their business processes.

"Business/IT Operations" alignment is properly performed as defined by ITSM/ITIL Best Practices, and as illustrated in my graphic below. Business and IT need to work as a intimate partnership to define, implement, deliver, and continually refine an optimized Service Portfolio at contracted service levels and an established and predictable cost point. Again, SOA is simply a technique that helps IT achieve operational excellence.

All other functions are internal to IT. The fact that requirements are fleshed out in an Agile fashion and constructed/deployed using a SOA strategy is meaningless to the Business Unit. They simply want IT to build the capability they need, adjust it when asked, and deliver it as expected.

<>

As a consumer and purchaser of various utilities (electricity, gas, cable, phone, water, etc) you don't need nor want to know the details of how the utility achieves scale economy or service resiliency or security or efficiency/utilization or adaptability or regulatory compliance. Well, okay, you and I by nature might be curious and like to know how these things work. But, in general, exposing the internal details of how a Service Delivery Platform is constructed is, IMHO, counter-productive to the Business/IT conversation and partnership. Some curious BU stakeholders will likely want to understand and even attempt to influence your model (eg: buy EMC, use .NET, etc). But that kind of inquiry can expose dysfunction and introduce inefficiency in the model. You don't tell Pacific Power to buy GE turbines or supply power at 62 hertz, unless you want to pay extraordinary fees for your own custom power plant.

I strongly believe in the principles of Agile development and architecture. Clearly the days of throwing a fixed requirements document over the wall are over. Business Units, IT Operations, and IT Development all must work together in a healthy partnership focused on continuous business process optimization and refinement. However, in my opinion, the true value of SOA is in the benefits it delivers to the internal IT function w.r.t. scale economy, resiliency, efficiency, adaptability, etc. Business Units don't need nor want to know about SOA... they simply have (frequently changing) requirements and expectations.

Bottom line: SOA is a architectural style/technique that IT Shops will employ to quickly respond to changing service level demands, while operating IT as an efficient adaptable business with an ability to tap into (integrate with) external/outsourced partners (blog on Coase's Law).

John - I respectfully invite a reply.

March 14, 2005 07:25 AM EST Permalink

Deciphering the Data Cent

Puzzles

Disclaimer: I've consulted at all but one of these firms, and can tell you that this puzzle's solution does not generally describe actual purchase decisions or deployment standards at these firms. It's just a puzzle.

Your job is to figure out each firm's preferred choice of servers and storage.

1. WalMart, the firm using SGI servers, and the firm with Hitachi storage, are all market leaders.
2. Newsflash: CNN uses IBM servers for their back-end processing.
3. Neither AT&T nor Disney have any AMCC storage in their data centers.
4. The firm still using SGI uses neither NetApp nor EMC storage.
5. Hitachi and NetApp were eclipsed on the Sun shop's preferred storage vendor list.
   
6. Both WalMart and Google have kicked EMC out of their centers.
7. Disney did not entertain the proposal to start using Fujitsu servers.
8. A search of Google's operations shows no sign of HP or SGI.
9. A Disney "balance-of-trade" agreement left LSI Logic without a ticket.
10. The firm using Fujitsu servers does not permit NetApp based storage in production.
11. Google is just starting to consider bringing in Hitachi storage arrays.

The following graphic is NOT the solution, just a randomly ordered list of related logos. It provides no useful information beyond the statements above (I just like to include a graphic with each puzzle). Good luck.

In case you give up, here is the solution:
http://blogs.sun.com/roller/resources/dcb/SOLUTION_Deciphering_Datacenter.html

March 14, 2005 07:20 AM EST Permalink

ITSM: Transforming IT

Computers

Here are two recent letters I sent to customers following workshops designed to map out a strategy to transform their IT organization thru the assessment of their people, processes, and technologies and the application of best practices. I thought that these might be beneficial to others who are attempting to do likewise. There are no great pearls of wisdom here, but it might get you thinking about having the conversation. ITSM = IT Service Management.

One client is attempting to synthesize several frameworks (ITIL, Sigma, ISO, and CMM-I) into a multi-year strategy to uplevel their operational capability. They asked for a mapping between ISO and ITIL, to which I replied in the 2nd letter (below).

Hi <--->,

I'm glad to see you are moving forward with this. As we mentioned during our workshop, some clients choose to perform the SunTONE assessment by themselves. Others seek assistance from Sun or a partner. Still others do both... performing an informal survey themselves and then requesting a formal evaluation from Sun. Either way, since I'm just down the street from you, I would like to keep tabs on your efforts and help ensure you get the assistance you need and the results you desire. If you find there are areas that you'd like to target for improvement, I can also help suggest services and/or technologies and/or best practices that will help improve your "score". Of course, it isn't about the score - but a firm's ability to deliver a quality service and experience that meets documented SLOs at a desired level of security and cost.

As I've mentioned, your operational capability is already (it appears) at a higher state of maturity than most. A SunTONE "stamp" will certify this capability and is a badge of honor. You'll join hundreds of other firms that have attained this status, and will differentiate yourselves from the other hosting centers.

If you have a standing meeting to discuss status and actions and gaps associated with this effort, and if you think I could add value to this meeting, I would be more than happy to attend and provide insights and suggestions where appropriate.

Hi <--->,

I'm more of a Sigma guy than an ISO guy... But from my investigation of ISO, it seems clear that a clean mapping exists between ISO and Sigma. These are initiatives to create and document and control processes to ensure a high degree of quality and predictability and continuous improvement/refinement. These are wonderful tools to ensure a process continues to be aligned with expectations and goals, and is as efficient as possible.

ITIL and SunTONE, on the other hand, DEFINE best practices and processes.

See the difference? ITIL is a set of practices/processes, whereas Sigma and ISO are mechanisms to ensure any process is (and continues to be) optimized.

So, in that sense, they are HIGHLY complementary, but orthogonal. I don't believe there is overlap or mapping between ISO and ITIL. You really need both the processes (ITIL) and the means to define and measure and analyze and implement and control (Sigma/ISO) those processes.

Note that both ITIL and Sigma/ISO are systemic/intrusive frameworks that, if done right, will infiltrate the whole organization and will be embraced and promoted from the highest levels. It is a culture change that takes more than a training campaign, MBOs, and a tiger team. You already know this, but many clients fail because they are not prepared to endure the multi-year evolution that this kind of change requires. But, for those that succeed, there are great rewards all along the way... incremental quick-hit benefits that don't require huge time or resource investments.

Many IT shops, I believe, will be outsourced and/or be "consolidated" over the next few years because they can not control their costs, security, and service levels. ITIL+Sigma/ISO is the path to survival and excellence.

Hope this helps!!

March 11, 2005 04:50 PM EST Permalink

Java Jingle

Computers

 Java Jingle from 1997:  http://blogs.sun.com/roller/resources/dcb/Java.mp3

I think Sun employees wrote and recorded that song. Anyone recall who? A verse near the end states: "Nobody can tell you what the future may bring...". Well, that was 8 years ago. Check this out!

As Java technology enters its 10th year, the Java Brand is a one of the most powerful technology brands on the planet. You'll see it on your Java powered mobile phone from Sony Ericcson, Motorola, etc or your Palm PDA, on a variety of new PCs from the factory, built into various printers from Ricoh, baked into mobile games, and a part of slew of websites from our partners like Borland, Oracle, and others. Java technology is on over 2 Billion devices and counting!

The Facts
In our most recent study we found that 86% of consumers and 100% of developers and IT recognize the Java brand. In addition we have seen the association of Java and Sun grow by 15% year over year. Over 80% of Developers and IT professionals know that Java comes from Sun. In addition 1 in 3 consumers will buy a product with the Java brand over a comparable product, this is up from 1 in 5 just a year ago.  Java.com just blew past 10 Million visitors per month, which is more visitors than Nintendo.com, Wired.com, Playstation.com, Time.com, Businessweek.com, and many others. Here are some facts and figures:

• 2 million downloads of J2EE 1.4 - the most popular release ever!
• 4.5 million Java developers, up 7% from June 2004
• 2 billion Java-enabled devices, up 14% from June 2004
• 750M Java Cards, up 25% from June 2004
• 579M phones, up 65% from June 2004
• 650M PCs, up 8% from June 2004

March 11, 2005 02:09 PM EST Permalink

Power Hungry Grids!!

Computers

I find it ironic that our industry uses Power Generation and Distribution Grids as a metaphor to describe the utility based computing model that is being promoted by vendors and demanded by an increasing number of customers. Actually, it is a reasonable and appropriate analogy. You don't build your own unique power generator for your home or business, and you don't hire a Chief Electrical Officer. Instead you plug into the Power Grid(s)... and leverage standards and scale economics and the variable cost structure of a reliable shared service provider for which you pay for what you consume at a predictable cost per unit. Being a commodity adhering to standards, you can easily switch providers with little or no impact to your operation. You demand a level of service quality, and know what you are willing to pay for that service.

I find it ironic simply because it will take a main artery from the Power Grid to, well, power the Compute Grids being designed. There are plans on drawing boards to increase the compute density of future servers such that a standard 19" datacenter rack will (fully populated with the most dense compute servers) consume up to 25KW of power!! That's huge. Consider a data center floor filled with these racks. You can imagine the engineering challenges associated with extracting that much heat from these blast furnaces. And then, of course, it's up to the datacenter to do something with that all that heat. One customer measured hurricane force chilled air speeds underneath their raised floor tiles! To make matters worse, according to p.20 of this report (see the table below), computer equipment accounts for less than half of the power demand for a typical data center.

The good news is that you'll have an unprecedented amount of compute power on each floor tile, so in theory, you won't need as many racks. Of course, we all know that the demand for compute capability exceeds the supply. On the other hand, the ultimate realization of the utility model suggests that you might not even have your own datacenter. Like your gas, water, electricity, cable, and phone services, the cost of the building, of powering, cooling, and administering the equipment, of security, insurance, disaster recovery, etc, will all be absorbed by the utility provider. You simply pay for the service at a known rate per unit of consumption.

That sure sounds great in theory (unless you are the Chief Integration Officer, or Chief Infrastructure Officer). It'll be fun to watch this play out. And watch IT earn the title: "Information Technology".

March 11, 2005 10:03 AM EST Permalink