by Anonymous Coward writes:
on Wednesday November 29, 2000 @06:20PM (#592993)
I put in 7 years there before become a professional Linux kernel hacker..;) So sit back, I've had a few beers and I'm pretty animated so I'll walk you through this mess.
It work's like this. IBM is amazingly big, hard to believe how big, they make, build, design, construct, develop, support an absolutley huge number of products. I believe they have offices in all but 2 or 3 countries around the world, they actually do development in something like over 100 countries. It's insane. It all works because they trust their managment and they distribute stuff.
Managment is given "ownership" over various things, the higher up the manager the more he owns. Gerstner runs the whole show, under him are a bunch of senior VPs, maybe 20 or so and the company is divided up in to divisions with a senior VP at the head of each one. Like microelectric, print, and stoarge make up the "technology division." Then those divisions are further divided up into companies and subdivisions with a junior VP running/owning each one. Those junior VPs have developement staffs with managment staffs that own various parts of the devlopment group, sometimes marketing with a staff, always sales with staff, support staffs, etc. They are all independant, and given ownership of their operations. If they run their part well then they are rewarded, if not then a new manager get's to run it and the old manager does something else of has his wings clipped or leaves.
Then there is the concept of personal business commitments, or "PBCs" to use the IBMeese. Your senior VP will set goals, like he is going to bring in $6billion and create a new revenue stream worth $1billion more than last year and develop a new line of widgets and a new brand of jujus. The junior VPs, depending on which branch they are in take on parts of those PBCs and further refine then, so Jr VP X is going to enhance last year's widget to create a new line and sell xyz more of them at q% more profit to make $250million in new revenue while maintaining what he did last year. Sometimes the more senior manager will rewrite or "adjust" the junior manager's PBCs, so if $250million isn't enough he might bump it to $400million, there is supposed to be discussion on this but we've all been in relationships with people above us and know how that goes there are some tricks though. That trickles through 4 more layers down to a developer who then commits to learn java, fix all field reported bugs in less than a week, and jack around less or something like that; and his manager has to agree to those and will change them if need be, again there is supposed to be discussion but it's usually in 1 direction; "you're going to learn java and c#." (If you play the game right, that's what happens, if you're foolish you'll get strapped with something impossible.) So there is this nice hierarchy of commitment and it's a lot like the game of telephone, your VP is focused on bucks and by the time it get's to the worker bees it's about "leveraging java" or something, each manager in between makes adjustments so that he can do what he feels is the most likely to help him meet his objectives, and there is resistence you don't want to be 4 layers down and commit to bringing in $1billion because it's not going to happen and you'll be penalized for it. To do Linux or anything new you need to have two or three consecutive layers agree that it will help them out.
So for example, my first line manager owned 5 or 6 applications and the staff who developed them, his performance was based on how well we meet our PBCs which was directly tied to the product. the performance measurement then trickles back up the chain to some senior VP who sees a number on the bottom of a ledger which is hopefully black and something as big or bigger than the one he made up 12 months earlier. If that happens then all is good and everybody get's a nice bonus in February and a big raise in March, if not then he get's pissed off and whips are cracked to find out who didn't meet their objectives and caused "the team to fail." As you can see, there is lot's of translation going on and it really does get back to the worker bees. Because they didn't leverage java or fix bugs fast enough or something the team performed poorly, so pressure is applied, they try to focus on what it is that they do well and not risk and change is slowed down by this process.
So where am I taking this train? There is a simple way to short circuit the problem, the bottom line is you're not going to sell $1billion more widgets this year becuase widgets are stupid and the market is flat and Yoyodyne in Sunnyvale is make a mojo that costs 1/10th what a widget does and does 5x more, plus it has a cool colorful case. So as the PBC trickle down affect happens, the more junior people start setting lower and lower goals for themselves, that way when the Senior VP get's pissed and starts cracking whips there is backlash. For some reason that I don't understand this works, if you say you will do 25x your manager can change it to 30x but if you say you're going to do "applepie" your manager can't really say "2applepie" and if the goals are made so that they are more easily measured then the work bee picks one that may be high but it's in his backyard and easy to achieve. The worker bees set goals like "I'm going to learn java and get to work by 10:00am every day" and he achieves those goals, with ease. His boss tries to apply pressure and he says "I learned java and I haven't been late once, in fact I've been here by 7:00 everyday. I've exceeded my goals and done my piece." and defuses the situation or softens the blow. The more popular tactic is to set goals that can't be understood (various quality assurance metrics and managment philosophy type goals are popular) or goals can't be quantified. I've had managers show me their PBCs and I couldn't understand half of them and I have a two advanced degrees...
There is also one more wildcard factor I call "the hand of God." If a senior VP get's a bug up his ass about something technical or specific, then it is done regardless of objectives, even if all his people have objectives that totally contradict it and that is usually the case. If Lou Gerstner decides that Linux is the hip thing and he wants to see it on a ThinkPad, then there is going to be a Thinkpad with "BlueHat Linux" on it and it is going to exist fast even if the Thinkpad people are all about windows. "Hand of God" is powerful in that it makes the company look fast it also has a negative affect of creating animosity. When you miss your PBC because you've been building Bluehat Linux, you're going to be pissed off and you're not going to do anything Linux for a while.
So how Linux steps in to this is kind of orthagonal. First, products are owned and if a product is to be ported, supported, etc.. then the team that owns it is going to do it. Why? because giving it to another group could potentially cause that group to fail to meet their goals or if they do then it is success that doesn't trickle up to your boss through you, the culture is to keep ahold of things until they are miserably out of date. Second, that team tries to set low goals for itself so it can achieve them and numerous times it has been shown that going cross platform is both hard to do, time consumnig, and almost never shows the kind of profit that trickles up in to the figures corporate is looking for. Then lastly becuse of the higher level ownership there are almost no parallels between products. For example, OS/2 was made by one division and PCs by another; from a consumer's point of view they go together but OS/2's success (think PBCs) has nothing to do with the PC's success because they are being measured by different metrics and realistically OS/2 would have initially hurt or been a risk for the sales of PCs the PC group wanted nothing to do with it.
Same is true with Linux. Db2 is made by one group and PCs another; Linux runs a lot of web servers and DB2 is a great database for web stuff so getting DB2 onto linux makes a lot of sense for them becuase it opens a market. On the otherhand 95+% of PCs sold run windows so to do well in that business and be part of the 95% you don't want to screw around with Linux. Never mind the fact that the PC is being sold as a server. The PC guys take bigger risk by trying to push linux when they can set modest goals (sell 5% more WindowsME PCs this year, which is actually pretty good and pretty tough to do) and achieve those with relative ease. Linux has really ended up on IBM PC hardware because of the "Hand of God" and it was done against resistence.
PowerPC is in the same boat, PowerPCs are made by microelectric which isn't anywhere near the division that PCs are in. They tried to put PowerPCs in to PCs for something like 45seconds before they dropped it because it wasn't going to be profitable. There was some HoG involvement and now the PowerPC has lost any reputation it could have possibly had with the PC people because they pissed away millions of dollars trying to make it work, it's not just a risk but they hate it because they were burned by it in the past. Don't expect to ever see IBM sell PowerPC based machines to the consumer.
It's the software / hardware rift that makes it the most complex, the success of software has nothing to do with the hardware and vice-versa. We all know they go together and even IBM knows that to some degree but when it comes time to do better than you did the year before you want more options and you don't want to take big risks. Ad because software and hardware are different divisions there isn't a lot of overlap
Don't get me completely wrong, there are cowboys in the mix and there are always going to be hackers and geeks in the fray. The most powerful tool the hacker at IBM has is skunkworks, he ports DB2 to Linux at home and then presents his managment with a new product "for free" and they will usually not turn that down. All the big press Linux items started that way, DB2, domino, S/390, etc. Good things can happen, and senior managment are always wild cards, the guy who is slated to follow Gerstner happens to be a Linux freak and so Linux has been getting a lot of pub and press but still is having trouble finding its way into core markets because it's still seen as a risk and he dictates things. The overall culture is to try to avoid risk though and that means don't change until it's too late. The PC group in particular is in an intensely competitive market and has an extremely difficult time meeting their objectives (they fail most of the time) and so getting Linux on to Thinkpads and that type of stuff isn't done unless someone from the top orders it and then it's never going to be carried on.
IBM is big and complex and has it's own culture. (Score:5)
It work's like this. IBM is amazingly big, hard to believe how big, they make, build, design, construct, develop, support an absolutley huge number of products. I believe they have offices in all but 2 or 3 countries around the world, they actually do development in something like over 100 countries. It's insane. It all works because they trust their managment and they distribute stuff.
Managment is given "ownership" over various things, the higher up the manager the more he owns. Gerstner runs the whole show, under him are a bunch of senior VPs, maybe 20 or so and the company is divided up in to divisions with a senior VP at the head of each one. Like microelectric, print, and stoarge make up the "technology division." Then those divisions are further divided up into companies and subdivisions with a junior VP running/owning each one. Those junior VPs have developement staffs with managment staffs that own various parts of the devlopment group, sometimes marketing with a staff, always sales with staff, support staffs, etc. They are all independant, and given ownership of their operations. If they run their part well then they are rewarded, if not then a new manager get's to run it and the old manager does something else of has his wings clipped or leaves.
Then there is the concept of personal business commitments, or "PBCs" to use the IBMeese. Your senior VP will set goals, like he is going to bring in $6billion and create a new revenue stream worth $1billion more than last year and develop a new line of widgets and a new brand of jujus. The junior VPs, depending on which branch they are in take on parts of those PBCs and further refine then, so Jr VP X is going to enhance last year's widget to create a new line and sell xyz more of them at q% more profit to make $250million in new revenue while maintaining what he did last year. Sometimes the more senior manager will rewrite or "adjust" the junior manager's PBCs, so if $250million isn't enough he might bump it to $400million, there is supposed to be discussion on this but we've all been in relationships with people above us and know how that goes there are some tricks though. That trickles through 4 more layers down to a developer who then commits to learn java, fix all field reported bugs in less than a week, and jack around less or something like that; and his manager has to agree to those and will change them if need be, again there is supposed to be discussion but it's usually in 1 direction; "you're going to learn java and c#." (If you play the game right, that's what happens, if you're foolish you'll get strapped with something impossible.) So there is this nice hierarchy of commitment and it's a lot like the game of telephone, your VP is focused on bucks and by the time it get's to the worker bees it's about "leveraging java" or something, each manager in between makes adjustments so that he can do what he feels is the most likely to help him meet his objectives, and there is resistence you don't want to be 4 layers down and commit to bringing in $1billion because it's not going to happen and you'll be penalized for it. To do Linux or anything new you need to have two or three consecutive layers agree that it will help them out.
So for example, my first line manager owned 5 or 6 applications and the staff who developed them, his performance was based on how well we meet our PBCs which was directly tied to the product. the performance measurement then trickles back up the chain to some senior VP who sees a number on the bottom of a ledger which is hopefully black and something as big or bigger than the one he made up 12 months earlier. If that happens then all is good and everybody get's a nice bonus in February and a big raise in March, if not then he get's pissed off and whips are cracked to find out who didn't meet their objectives and caused "the team to fail." As you can see, there is lot's of translation going on and it really does get back to the worker bees. Because they didn't leverage java or fix bugs fast enough or something the team performed poorly, so pressure is applied, they try to focus on what it is that they do well and not risk and change is slowed down by this process.
So where am I taking this train? There is a simple way to short circuit the problem, the bottom line is you're not going to sell $1billion more widgets this year becuase widgets are stupid and the market is flat and Yoyodyne in Sunnyvale is make a mojo that costs 1/10th what a widget does and does 5x more, plus it has a cool colorful case. So as the PBC trickle down affect happens, the more junior people start setting lower and lower goals for themselves, that way when the Senior VP get's pissed and starts cracking whips there is backlash. For some reason that I don't understand this works, if you say you will do 25x your manager can change it to 30x but if you say you're going to do "applepie" your manager can't really say "2applepie" and if the goals are made so that they are more easily measured then the work bee picks one that may be high but it's in his backyard and easy to achieve. The worker bees set goals like "I'm going to learn java and get to work by 10:00am every day" and he achieves those goals, with ease. His boss tries to apply pressure and he says "I learned java and I haven't been late once, in fact I've been here by 7:00 everyday. I've exceeded my goals and done my piece." and defuses the situation or softens the blow. The more popular tactic is to set goals that can't be understood (various quality assurance metrics and managment philosophy type goals are popular) or goals can't be quantified. I've had managers show me their PBCs and I couldn't understand half of them and I have a two advanced degrees...
There is also one more wildcard factor I call "the hand of God." If a senior VP get's a bug up his ass about something technical or specific, then it is done regardless of objectives, even if all his people have objectives that totally contradict it and that is usually the case. If Lou Gerstner decides that Linux is the hip thing and he wants to see it on a ThinkPad, then there is going to be a Thinkpad with "BlueHat Linux" on it and it is going to exist fast even if the Thinkpad people are all about windows. "Hand of God" is powerful in that it makes the company look fast it also has a negative affect of creating animosity. When you miss your PBC because you've been building Bluehat Linux, you're going to be pissed off and you're not going to do anything Linux for a while.
So how Linux steps in to this is kind of orthagonal. First, products are owned and if a product is to be ported, supported, etc.. then the team that owns it is going to do it. Why? because giving it to another group could potentially cause that group to fail to meet their goals or if they do then it is success that doesn't trickle up to your boss through you, the culture is to keep ahold of things until they are miserably out of date. Second, that team tries to set low goals for itself so it can achieve them and numerous times it has been shown that going cross platform is both hard to do, time consumnig, and almost never shows the kind of profit that trickles up in to the figures corporate is looking for. Then lastly becuse of the higher level ownership there are almost no parallels between products. For example, OS/2 was made by one division and PCs by another; from a consumer's point of view they go together but OS/2's success (think PBCs) has nothing to do with the PC's success because they are being measured by different metrics and realistically OS/2 would have initially hurt or been a risk for the sales of PCs the PC group wanted nothing to do with it.
Same is true with Linux. Db2 is made by one group and PCs another; Linux runs a lot of web servers and DB2 is a great database for web stuff so getting DB2 onto linux makes a lot of sense for them becuase it opens a market. On the otherhand 95+% of PCs sold run windows so to do well in that business and be part of the 95% you don't want to screw around with Linux. Never mind the fact that the PC is being sold as a server. The PC guys take bigger risk by trying to push linux when they can set modest goals (sell 5% more WindowsME PCs this year, which is actually pretty good and pretty tough to do) and achieve those with relative ease. Linux has really ended up on IBM PC hardware because of the "Hand of God" and it was done against resistence.
PowerPC is in the same boat, PowerPCs are made by microelectric which isn't anywhere near the division that PCs are in. They tried to put PowerPCs in to PCs for something like 45seconds before they dropped it because it wasn't going to be profitable. There was some HoG involvement and now the PowerPC has lost any reputation it could have possibly had with the PC people because they pissed away millions of dollars trying to make it work, it's not just a risk but they hate it because they were burned by it in the past. Don't expect to ever see IBM sell PowerPC based machines to the consumer.
It's the software / hardware rift that makes it the most complex, the success of software has nothing to do with the hardware and vice-versa. We all know they go together and even IBM knows that to some degree but when it comes time to do better than you did the year before you want more options and you don't want to take big risks. Ad because software and hardware are different divisions there isn't a lot of overlap
Don't get me completely wrong, there are cowboys in the mix and there are always going to be hackers and geeks in the fray. The most powerful tool the hacker at IBM has is skunkworks, he ports DB2 to Linux at home and then presents his managment with a new product "for free" and they will usually not turn that down. All the big press Linux items started that way, DB2, domino, S/390, etc. Good things can happen, and senior managment are always wild cards, the guy who is slated to follow Gerstner happens to be a Linux freak and so Linux has been getting a lot of pub and press but still is having trouble finding its way into core markets because it's still seen as a risk and he dictates things. The overall culture is to try to avoid risk though and that means don't change until it's too late. The PC group in particular is in an intensely competitive market and has an extremely difficult time meeting their objectives (they fail most of the time) and so getting Linux on to Thinkpads and that type of stuff isn't done unless someone from the top orders it and then it's never going to be carried on.