On Being Open

New BMC Open Source management tool for MF Linux

stevecarl — Thu, 09 Oct 2008 23:55:47 GMT

Hi all

Been a while since I posted anything here, and for a reason: Been working on, among other things, getting a new tool called VMLMAT out the door.

VMLMAT is Open Source (BSD Licence), and available at sourceforge:

http://vmlmat.wiki.sourceforge.net/

I just posted about it over at TalkBMC here:

http://talk.bmc.com/blogs/blog-carl/steve-carl/VMLMAT-BMC-Open-Source-tool

VMLMAT is a pretty specific managment tool: It is about managing and archiving Linux images on the mainframe. it is implemented in Open Sourcetechnologies like PHP, Apachie, Samba, and the like.

Full disclosure: I did not write this tool: I was the manager of R&D support at the time and I commisioned someone on my team named Ron Michael to write it. When Whurley came a calling looking for technology inside BMC that could be Open Sourced, this tool was a natural. Ron agreed, and the rest is history... more or less.

Had to tell someone who would understand....

Being Open. Being Inclusive. Being Human.

stevecarl — Sun, 24 Feb 2008 02:24:41 GMT

Over the years I have had somewhat mixed results reaching people with various points in my blogs. I usually work pretty hard to set the stage for a position I am about to take. In some cases, I have written an entire series of posts before coming to the main point. Not everything is easy to say. Not everything can be functionally decomposed into byte size morsels.

A recent example was a post I did over at TalkBMC. I tried to set up the framework of what I was looking at and why, but I came across some comments in another forum about my post and it was clear that I had not succeeded. At least not for the people inspired enough to respond in the forum. The comments were all over the place: Anywhere from "So What?"(which is fine: The subject was specific enough that if one was not affected the post was probably not interesting) to one like "What a loser, he spent all his time saying this and that". i wondered if they had read the post I had written at all. I went back and re-read it just to be sure. Nope. I had not said what they thought I had. Yet other people had gotten the point. How had that happened?. The ones with the negative reactions often focused on the bad things, ignored the caveats and background info, and reacted. That was just what pressed their buttons.

For some people, The feeling was I was calling their baby ugly. No.. More like:

"Hi. I like your car" I bet that gets good gas mileage. Hey, I also like your iPhone, the case you have, the headphones, your nifty hairstyle, awsome sunglasses, and those shoes you have on. Wherever did you get those? Whats up with the socks though?"

"Socks! Why do you hate on my socks? These are my favorite socks. There are two of them and they fit my feet and ... and.... and... what is so great about your socks anyway? Have to chase that person far to get them?"

When I think about all the open source projects out there, and even just the ones here that are related to monitoring and managing systems, I sometimes wonder at all the energy that has led people to go off into so many different directions to solve the same problems. Part of it is simply that there is more than one way to solve almost anything of course. I am pretty sure that another aspect is human emotional needs entering into and interferring with the other goals of what is otherwise supposed to be an inclusive project. There seems to be a common trajectory (or maybe small set of possible trajectories) to Open Source projects. and it looks something like this:

Someone gets an idea, or has a need and works to create a solution to it.
They want to share their idea / solution with others
If it catches on, sooner or later it hits critical mass and enough people are involved that they start to pull different ways. They may do this based on their own needs, their own ideas not being accepted by the core group, or suddenly thinking they have some advantage heading off in their own way. Schism. Division.
Soon there are multiple solutions to the same problem. There are probably good things about each one, and also probably problems with each approach.
Maturity is reached where things come back together. Or Interoperate. Or things start to die.

Here are three examples I am thinking of:

Compiz / Beryl: They split, and Compiz created Beryl in the same sense that certain greek gods used to be born, spring from the cracked skull of a parent. But now they are back together than Compis Fusion is the result. We all won on that one.

Gnome / KDE: They could not agree on a core tool set and they split, with Gnome springing from to propritary (at the time) clutches of Trolltech. They did not interoperate well at first, despite both being X applications. Later they came to a peace and found ways to let each others applications run inside their environments. Again we all won.

Evolution / KDE PIM: Evolution created the Connector. The Connector let one attach to MS Exchange. KDE updated to Kontact to do a similar trick, but stopped half way through development. Connector was open sourced, but KDE did not adopt any part of the Connector code. Kontact stayed in its half way state, while KDE focused on other things like look and feel and integration of KDE and a new major release of the desktop. No one won there. There is still only one native application way of getting at your calendar from Linux.

But I bet you could point at examples of the same thing inside of Open Systems management. What are they, and how do we head them off at the pass? How do we make sure we work together now?

The VW Beetle Principle

stevecarl — Thu, 17 Jan 2008 03:57:09 GMT

First off… I admit it. I totally made that principal up. I needed a sort of reverse (if dated) example as a counterpoint for Whurleys “Bugatti Principle” post over at TalkBMC from last year:

http://talk.bmc.com/blogs/blog-whurley/whurley/the-bugatti-principal

The VW beetle would seem to be the antithesis in most ways, at least in principle and concept, from the Bugatti Veyron. Not exclusive. Not fast. In fact designed to have the absolute minimum required and still be a workable car for the road of its day. I had a 1965 Beetle convertible with a 1200cc, 40 horsepower engine (which I admit I hacked. I “dropped in” actually shoved up a 1600cc, 67 horsepower engine to replace the original mill. It felt like a race car in first gear)

The car was not fast. It got fairly good mileage for the day (about 27 MPG: Ludicrous by todays standards). At about 65 miles per hour the shape of the car made it tend to want to lift off the road, and depending on wind direction 65 to 70 was all she would do. After that the wind resistance overcame the engine. The Beetle looked round but was about as aerodynamic as a brick. With its flat pan shaped bottom (the reason it actually would float for bit. I watched mine float till it sank during a hurricane once), and its arched body shape, it more strongly resembled a poorly shaped wing than a brick, but the principlel was the same. Its aerodynamic drag coefficient was .48, where a wall is 1.0, and a Hummer is .7.

Everything about the car was the minimum in late 1930’s-1940’s design. The heat for winters came from the finned exhaust pipes to the front two cylders of the flat four. Air was take from the cooling fan and diverted over these fins, and then into uber-hot but tiny floor vents. Once it was warmed up in the cabin, It was a pretty pleasant place to be in the winter, but it took forever to get warm. Plastic laid near the tiny vent would melt. I know.

There were two moderately comfortable seats, a speedo, an AM radio, a glove box, a 10 or so gallon fuel tank, a full size spare, four 15 inch, skinny tires, and windows that rolled up and down with hand cranks. As basic as the beetle was I suppose you could go even more basic and delete the radio and the heater and the roll up windows and still have a useable car, but it was pretty basic. Very close to the smallest and fewest parts a car could be and still be a car and not something else… like a tricycle or a motorcycle. And yes, some Beetles have been made into both of those.

The Beetle was the low cost leader in the US in the 1950’s and early 1960’s: they predated Datsun and Toyota and all that. British Leyland cars were anything but inexpensive, especially once you factored in the Lucas Electrics (Prince of Darkness) repair bill. I remember seeing ads in the paper when I was kid for 1900.00 USD for the sedan. 1960’s money. Adjusting for inflation that is about 12,000 USD now. That looks about right too: A 12k car in the US is at the most basic end of the market, even if it is technically a much better car: more fuel efficient, less pollution, more crashworthy, AM/FM radio….

Related to Systems Management?

The Beetle was as basic a car as it could pretty much be and still be a car. Factored into that were all the things that it took to be a car, and all the costs that went into developing and manufacturing and updating the car. In fact the 1965 car looks positively modern when you compare it to the 1940’s prototypes. Those had mechanical flags rather than electric turn signals!

To support that car was a set of designers, engineers, QA testers, Sales and marketing folks, and all sorts of hidden costs like the cost of building and maintaining the steel stamping gear, the aluminum engine casting plant, and on and on. There was a minimum cost below which the Beetle could not cross and make any money. It had to make money. VW was and is a commercial car vendor.

Systems management is hard. On the high end (The Bugatti end), it takes a great deal of work to successfully manage large, complex, heterogenous systems and to do it well. Do it invisibly. Do it as if it were data tone.

At the same time, there is a lower limit below which a commercial vendor cannot easily cross. To maintain that manufacturing line and all the staff, the building, computers, and benefits. The travel budgets and comm gear. On and on. To go below a critical size is to move into the realm of unprofitability. No public company can be unprofitable for long before they board fires the executives and find some others that will make it profitable.

Size Works at Least Two Ways

One kind of “lower limit” in systems management would be number of managed nodes. A management system that is designed to “boil the ocean”, I.E. manage huge environments makes no sense below a certain level. How far a management system can scale down when it is designed to work for say 100,000 or 500,000 managed nodes. That puppy is Bugatti’ed to death. It is tested, built, designed, engineered, scaled, and everything is oriented around making sure that it can “go as fast” as is advertised. It would almost be silly to use it to manage, say, 10 systems. First off, 10 systems don’t even need that kind of management. If they are all in one place, you can sit and see all their screens at once!

Another kind of scale issue is the type of managed node itself. There are a zillion devices in this world, and IPv6 is going to make it so that everything on the planet sooner or later has an IP address. My fridge, my toaster, my stereo… everything. If an electron can run though it, the possibility is there that it might need an IP address someday.

All that diversity usually leads to a “Tower of Babel” though. As the wit once said, “The good thing about standards is that everyone has them”. Unless Open Standards are used, this is going to be a real problem for the widget in question.

On the one hand, out here at this edge is where a great deal of the innovation is going to be. The latest widget that sets the world on fire will come from here. Or Apple. ☺ Maybe I should have used the "MacBook Air Principal" here... But I digress.

While the widget is tiny, unless they are playing in a space where an Open Standard already exists (and if they are creating a category, it probably does not) then it may not be worth everyone supporting it I am not even talking about competition here, but time/effort to add support to an existing solution. This alleged world turning widget may come from a big vendor, a small vendor, or a child with the OLPC XO and a good idea. Whatever it is, while it is small… while it does not command enough market share to hit the lower limit of a management vendor to create a management tool for it, then its management is either created at the same time as the widget by the same creator, or by the first person who needed to manage it so badly that they gave up waiting for the vendors and wrote it themselves.

If it Were Easy, Anyone Could Do It

All of the above listed conundrums and complexities are just to underline the point that systems management is not easy. Even at the lowest end of things, there are things that any solution has to have, and has to deal with. They are not always technical. But that takes me to….

The Honda Fit (Jazz) Principle

Because my new car is a Honda Fit of course. I sit in the car and I marvel at it sometimes. How this tiny econobox (SuperMini actually) has benefited from all that came before it.

I have owned many Hondas over the years. A 1984 Prelude, and 1986 Accord, a 1992 Integra, and now this 2008 Fit. This is the best car of the bunch. It is low emissions (number three best in 2007), gets fairly decent fuel economy (38 MPG best tank so far), and is relatively quiet and comfortable for such a small car. It is a littler faster than I was going for, but in the US they don’t give us the option of the smaller, even better fuel economy engines.

I see the lineage of all the other Hondas I have owned here. The design, the way it sounds, the way it handles, the materials choices, the control layouts. On and On. This car has benefitted from all the 100+ years of car design before it. The best shape for the wheels being round. Low drag wheels and tires being better. The rubber material in the tires being both “sticky” and long lived. The way the tread is now designed with a computer to siphon off the water on the road. Radial belts to lower rolling resistance. Light unsprung weight of the wheels helping handling. Disk brakes being better than drum. A radio that has more computer bits in it than my first TRS-80. The new little chips that are better than the old bigger chips which are better than the older transistors which are better than the old tubes. The current generation of plastics having a longer life, nicer touch, and being more recycle-able. This could be an infinite regression.. The car stands on the shoulders of all the cars that have come before it, and not just from Honda… and for that matter, not just of cars.

All computer programs, including systems management software are the same way. Innovations in version 1 of Patrol (for example, since I know that one) and the way it allowed system programming knowledge to be captured and automated were picked up and improved by others (and therefore made Patrol have to improve too). Having to hand build and hand install management toolsets giving way to automated discovery and automated install and provisioning servers. SNMP standards started out too loose, were tightened up and further defined, and later releases were far more useful that the first, and then most of the system management tools started being able to use them.

Layer upon layer, one thing building on another. Capability growing, along with complexity. Open Source / Open Standards / Open Frameworks, etc. makes all this layering and building and improving far easier, and far faster.

Scale and Moore’s Law

In my VW Beetle principal, the 1.9k 1965 Beetle and the 12k economy car of today show what happens as technology improves and time goes by. The least expensive car for sale today in the USA is a bit under 10k, inflation adjusted less expensive than the Beetle was in its day. That car is in every possible way ( no matter which car it actually is, but a quick google makes it look to be the Chevy Aveo at 9955.00 USD ) the technical superior of the 1965 car. In fact, Tata just introduced a 2,500 USD, 2 cylinder, 33 HP, 50 MPG car called the Nano:

http://www.boston.com/business/technology/articles/2008/01/11/indian_automaker_u nveils_worlds_least_expensive_car/

All the feature / function of the original Beetle at a quarter of its inflation adjusted price, and double its fuel economy.

Computers are even “worse”. While Moore’s law is not infinite, nor even really a “Law”, The over the counter cheapie computer I can buy for 200 USD right now is technically better than the 1000 dollar computer of five years ago more or less. Point being, an inexpensive computer can now manage quite a number of other computers. Where one used to have to install a steamship full of computers in the data center just to centralize all the data from the managed nodes, now it can be a raft.

At the same time, not all the complexity is dropping out of the equation. Sure, the number of computers needed for management dropped, but unless there is a hugely aggressive turnover in the data center to match it, the number of managed nodes probably grows over time, and the complexity of the managed nodes is probably is growing too. Can you say ‘Virtualization”? How about “Vista has 20 million lines of code in it”?

Don’t even get me started on analogies about re-inventing wheels.

While there are incredible things one has to do to build a Bugatti, and certain other scale problems like high end systems management, especially when integrated into overall ITIL solutions, there are also a certain number of minimum things one has to do. That is where being O/open really helps.

The steam train becomes the first car becomes the Model T becomes the Beetle becomes the Fit. I might have skipped a few steps along the way there…

I think I have tortured this analogy enough for one day. See you at BarCampESM this weekend!