Are contemporary rocket computer systems necessary for economical reusability? As I understand it, rocket launch costs stagnated for decades due to a lack of price competition stemming from the high initial capital costs involved in developing new rocket designs rather than us hitting a performance ceiling.
Eyeballing the chart here it looks like computers in 1995 were about 2-3 OOMs worse than computers in 2015. Payload capacity of Falcon 9 is measured in 1000s of kg. So you could cut off some of that payload capacity and have a 5000 kg computer in 1995. Would that be able to do the work of whatever computer they used in 2015? Well, how many kg did those computers weigh? More than 5kg? This thread has some info but no numbers.
idk, seems probable to me that computers weren’t the limiting factor.
It took quite a few attempts to get the Falcon 9 to land correctly. A lot of the work to get it to land correctly like involved getting the computer models right. Everything that’s involved in computer modeling got better over the 20 years.
Hardware in the rocket got cheaper. Compute for the offline models got cheaper. On the physics side, we likely learned lessons about how to model the airflow and other factors better.
Dragon runs a special version of Linux that’s optimized for low latency. It’s unclear to me whether a low-latency OS that would do the job was around in 1995.
I don’t know anything about the requirements for the Falcon or Dragon computer systems, but I do know that in 1995, real-time operating systems had existed for many years. At that time, computers that I would guess (maybe wrongly) would be adequate for the task weighed in at 10s of kilograms. Of course, the ones I worked with weren’t built to endure the rigors of space flight. Still, I would guess that the 1995 level of computer technology would not have been a show-stopper for such rockets being built, though the computing requirement would have added more to the cost than today.
But I wonder whether advances in materials science since 1995 might be crucial?
It’s a ballpark number, anything from 100kg+ would have made a 90s equivalent too heavy. The difference compared to consumer computers is expected once you realize it includes multiple interlinked triply redundancy systems and the wiring, mounting structure, power supply system, shielding, etc. for all that.
Would you consider the space shuttle doomed from the start then? Even without bureaucratic mismanagement, legislative interference, and persistent budget cuts? The market for rocket development in the 80s and 90s seems hardly optimal. You had OTRAG crushed by political pressure, the space shuttle project heavily interfered with, and Buran’s development halted by the collapse of the Soviet Union. A global launch market didn’t really even emerge until the 2000s.
As a broader point, even if you chalk up the nonexistence of economically competitive partially reusable rockets to Moore’s law, that still leaves an apparent gap in the development of more cost-effective expendable systems. Launch prices stagnated from the early 1970s until the 2000s. Surely expendable rockets in the 70s were not already as optimized as possible without 2000s computers.
Would you consider the space shuttle doomed from the start then?
Yes, but not primarily because of computer performance issues. The gliding descent profile, along with the rest of the operations, could have been entirely manually flown or manually controlled by ground based staff if necessary. Though that would have been economically less efficient.
It simply was too oversized and heavy for the vast majority of projected, and actual, missions.
Even without bureaucratic mismanagement, legislative interference, and persistent budget cuts?
It might have had much better economics if they had enough legislative support to approve the initial efficient designs, without needing to rely on the backing of the Air Force, who demanded nearly all of the cost growth as their condition.
Surely expendable rockets in the 70s were not already as optimized as possible without 2000s computers.
They were, some Soviet engine design from the 70s were the best for their niche until the late 2010s.
Manufacturing improvements a la SpaceX might have been possible in the 90s but at a much steeper price, as they were very niche and pricy techniques back then, so they wouldn’t have made expendable rockets any cheaper, unless ordered in huge quantities.
Hmm, I should rewrite the Falcon 9 sentence to clarify my intent. I meant to express that more affordable rockets were possible in the 90s compared to what existed, rather than that the F9 exactly was possible in the 90s.
They were, some Soviet engine design from the 70s were the best for their niche until the late 2010s.
Given that the Soviet Union collapsed soon after and that no competitive international launch market really began to emerge until the 2000s this isn’t surprising. There was no incentive to improve. Moreover, engines are just one component of the rocket launch cost equation.
The technical problems leading to high space launch costs have been identified and cures proposed, but the long delay until the recent reduction in launch costs suggests that cultural and institutional barriers have hindered implementing potential technical improvements.
One study suggested that the record low cost of the Saturn V could be reduced by a factor of 5, to a cost similar to the Falcon Heavy.
The fundamental cause of the past high commercial launch cost seems to be lack of competition. The US launch industry has been a monopoly, the United Launch Alliance (ULA), and its main customer has been the US government, NASA and the military, which need high reliability and had little incentive to exert cost pressure. The ULA lost most of the commercial market to Russia and Arianespace which are also heavily subsidized by their governments
In 2010, NASA compared SpaceX’s cost to develop the Falcon 9 to the cost NASA’s models predicted using the traditional cost-plus-fee method. Using the NASA-AF Cost Model (NAFCOM), NASA estimated that it would have cost NASA $1,383 million to develop these systems using traditional contracting. The estimated SpaceX cost was $443 million, a 68% reduction from the traditional approach.
Until recently, virtually all major players in this space (heh) were monopolies, whether public or private. These organizations had little incentive to improve and were known to be highly inefficient. Why assume the Soviets reached a magical price floor that was impassable prior to the 2010s?
And what incentives to improve could there have been?
If you mean on some alternate earth where human incentive structures are different and they had some Apollo sized project going on in their equivalent of our 90s, sure it probably would have been feasible to drive down the price of expendable rockets by a factor of 2 or 3 at their equivalent of 90s technology. Assuming they made it in huge quantities and the opportunity cost of capital was zero.
But nobody would have taken a fixed price contract to build rockets in our version of Earth, actual U.S., circa 1990s, without a huge profit margin built into it because rocket manufacturers also have access to accountants and actuaries, etc., who can price out possible risks. It didn’t help that there really only was one or two companies willing to invest capital into doing so. Which precluded the possibility of selecting a lower bid.
And no one else in the U.S. wanted to invest capital to establish a third manufacturer for the reasons described above.
Which is why the Pentagon hasn’t moved entirely to fixed price contracts, because for many systems there’s literally no competition for their business, so it would probably increase the price over cost-plus contracts since corporations have to borrow at higher interest rates for debt then the government can.
But nobody would have taken a fixed price contract to build rockets in our version of Earth, actual U.S., circa 1990s, without a huge profit margin built into it because rocket manufacturers also have access to accountants and actuaries, etc., who can price out possible risks.
It’s okay to grant them a huge profit margin with priced-out risks. That way, if they manage to reduce their costs, they make a bigger profit.
it would probably increase the price over cost-plus contracts since corporations have to borrow at higher interest rates for debt then the government can.
The government essentially guaranteeing the loans of companies does reduce the interest rate on the debt, because it means that if the project fails the government pays for the losses.
It also means that the chance that the project succeeds at its initial price is less because the company has no incentive to stay within the budget.
Right, so the upfront sticker price of a fixed price contract to build rockets in the 90s would have been much higher then that of a cost-plus contract.
Maybe after you include the delays, overruns, etc., it would turn out to be a lower price. But NASA didn’t need the votes at some future date, they needed the votes at the time of approval, in order for the project to happen.
Thus it would never get past Congress unless somehow NASA could guarantee that the congressmen voting for it would still be in power to benefit from the possible future savings, which is impossible in a democracy.
The problem is a mix of economic illiteracy of congressmen and corruption.
If we would make more fixed-prize contracts you could start attacking politicians who make cost-plus contracts on the basis that they cost the tax-payer a lot of money. If you try to enforce a standard of “any politician who makes cost-plus contracts that then run above budget is to blame for that”, you could shift the system to be more productive.
You write newspaper articles that blame the responsible politicians. You say that their actions resulted in a lot of wasted money and talk about how they took campaign donations from interests that profit from the government being responsible for the losses instead of private industry.
Elon Musk pushed for more space contracts to go fixed-price. In both space and defense, you could research who’s responsible for moving things in the right direction and who blocked it and resulting in wasted money.
If someone at Vox would decide they want to do something good for the world they could do it.
I don’t see how this could overcome the counter-efforts of those who currently benefit from cost-plus contracts. They after all have a lot more to lose, individually, then a society of several hundred million, where the per person costs may be a couple hundred dollars total in any given year.
Getting enough votes and maintaining voting discipline to enforce any standard at all is incredibly tough in the U.S. elections systems.
If a project goes fails and you write about how John is responsible for wasting a lot of tax-payers money because John decided to to a cost-plus contract instead a fixed-price contract, it’s hard to argue that John isn’t blameworthy if you can’t say who’s supposed to get the blame.
Avoiding opportunities to get blamed is a very strong motivator for many politicians.
What it takes, is enough reform-minded journalists who are willing to consistently talk about it for 1-2 decades.
This would work if it was the only contentious topic at stake during an election. However in reality, given recent trends, there will likely be dozens of hot button topics at stake and only a few viable candidates, and virtually all other topics carry more emotional appeal, and more motivated voting blocs, then fixed-price contracting standards.
It seems exceedingly unlikely that this issue would get enough oxygen for it to be decisive in selecting any elected candidate.
Are contemporary rocket computer systems necessary for economical reusability? As I understand it, rocket launch costs stagnated for decades due to a lack of price competition stemming from the high initial capital costs involved in developing new rocket designs rather than us hitting a performance ceiling.
Yes, computers of sufficient performance are necessary for economic reusability, most importantly for the automated landings.
How large (in kg, or in FLOPS) are the computers on the Falcon 9 today?
In kg terms, less than 0.1% of the total launch mass of the Full thrust variant (549 t).
Eyeballing the chart here it looks like computers in 1995 were about 2-3 OOMs worse than computers in 2015. Payload capacity of Falcon 9 is measured in 1000s of kg. So you could cut off some of that payload capacity and have a 5000 kg computer in 1995. Would that be able to do the work of whatever computer they used in 2015? Well, how many kg did those computers weigh? More than 5kg? This thread has some info but no numbers.
idk, seems probable to me that computers weren’t the limiting factor.
It took quite a few attempts to get the Falcon 9 to land correctly. A lot of the work to get it to land correctly like involved getting the computer models right. Everything that’s involved in computer modeling got better over the 20 years.
Hardware in the rocket got cheaper. Compute for the offline models got cheaper. On the physics side, we likely learned lessons about how to model the airflow and other factors better.
Dragon runs a special version of Linux that’s optimized for low latency. It’s unclear to me whether a low-latency OS that would do the job was around in 1995.
I don’t know anything about the requirements for the Falcon or Dragon computer systems, but I do know that in 1995, real-time operating systems had existed for many years. At that time, computers that I would guess (maybe wrongly) would be adequate for the task weighed in at 10s of kilograms. Of course, the ones I worked with weren’t built to endure the rigors of space flight. Still, I would guess that the 1995 level of computer technology would not have been a show-stopper for such rockets being built, though the computing requirement would have added more to the cost than today.
But I wonder whether advances in materials science since 1995 might be crucial?
Are the necessary algorithms parallelizable under the strict latency requirements?
2 OOMs would translate into a 50 000 kg computer system in 1995 v. a 500 kg computer system in 2015, not 5000 kg.
The computer on a 2015 Falcon 9 weighed 500kg? Really?
It’s a ballpark number, anything from 100kg+ would have made a 90s equivalent too heavy. The difference compared to consumer computers is expected once you realize it includes multiple interlinked triply redundancy systems and the wiring, mounting structure, power supply system, shielding, etc. for all that.
A100′s weigh less than 4kg. Three of them would be 12kg. I find it hard to believe that Falcon 9′s use 100kg worth of computer.
Would you consider the space shuttle doomed from the start then? Even without bureaucratic mismanagement, legislative interference, and persistent budget cuts? The market for rocket development in the 80s and 90s seems hardly optimal. You had OTRAG crushed by political pressure, the space shuttle project heavily interfered with, and Buran’s development halted by the collapse of the Soviet Union. A global launch market didn’t really even emerge until the 2000s.
As a broader point, even if you chalk up the nonexistence of economically competitive partially reusable rockets to Moore’s law, that still leaves an apparent gap in the development of more cost-effective expendable systems. Launch prices stagnated from the early 1970s until the 2000s. Surely expendable rockets in the 70s were not already as optimized as possible without 2000s computers.
Yes, but not primarily because of computer performance issues. The gliding descent profile, along with the rest of the operations, could have been entirely manually flown or manually controlled by ground based staff if necessary. Though that would have been economically less efficient.
It simply was too oversized and heavy for the vast majority of projected, and actual, missions.
It might have had much better economics if they had enough legislative support to approve the initial efficient designs, without needing to rely on the backing of the Air Force, who demanded nearly all of the cost growth as their condition.
They were, some Soviet engine design from the 70s were the best for their niche until the late 2010s.
Manufacturing improvements a la SpaceX might have been possible in the 90s but at a much steeper price, as they were very niche and pricy techniques back then, so they wouldn’t have made expendable rockets any cheaper, unless ordered in huge quantities.
Hmm, I should rewrite the Falcon 9 sentence to clarify my intent. I meant to express that more affordable rockets were possible in the 90s compared to what existed, rather than that the F9 exactly was possible in the 90s.
Given that the Soviet Union collapsed soon after and that no competitive international launch market really began to emerge until the 2000s this isn’t surprising. There was no incentive to improve. Moreover, engines are just one component of the rocket launch cost equation.
From NASA:
Until recently, virtually all major players in this space (heh) were monopolies, whether public or private. These organizations had little incentive to improve and were known to be highly inefficient. Why assume the Soviets reached a magical price floor that was impassable prior to the 2010s?
And what incentives to improve could there have been?
If you mean on some alternate earth where human incentive structures are different and they had some Apollo sized project going on in their equivalent of our 90s, sure it probably would have been feasible to drive down the price of expendable rockets by a factor of 2 or 3 at their equivalent of 90s technology. Assuming they made it in huge quantities and the opportunity cost of capital was zero.
NASA commissioned rockets under cost-plus contracts. Those contracts don’t produce incentives for the manufacturers of the rockets to cut costs.
If you want lower costs you need manufacturers to compete by making fixed-price bids.
But nobody would have taken a fixed price contract to build rockets in our version of Earth, actual U.S., circa 1990s, without a huge profit margin built into it because rocket manufacturers also have access to accountants and actuaries, etc., who can price out possible risks. It didn’t help that there really only was one or two companies willing to invest capital into doing so. Which precluded the possibility of selecting a lower bid.
And no one else in the U.S. wanted to invest capital to establish a third manufacturer for the reasons described above.
Which is why the Pentagon hasn’t moved entirely to fixed price contracts, because for many systems there’s literally no competition for their business, so it would probably increase the price over cost-plus contracts since corporations have to borrow at higher interest rates for debt then the government can.
It’s okay to grant them a huge profit margin with priced-out risks. That way, if they manage to reduce their costs, they make a bigger profit.
The government essentially guaranteeing the loans of companies does reduce the interest rate on the debt, because it means that if the project fails the government pays for the losses.
It also means that the chance that the project succeeds at its initial price is less because the company has no incentive to stay within the budget.
Right, so the upfront sticker price of a fixed price contract to build rockets in the 90s would have been much higher then that of a cost-plus contract.
Maybe after you include the delays, overruns, etc., it would turn out to be a lower price. But NASA didn’t need the votes at some future date, they needed the votes at the time of approval, in order for the project to happen.
Thus it would never get past Congress unless somehow NASA could guarantee that the congressmen voting for it would still be in power to benefit from the possible future savings, which is impossible in a democracy.
The problem is a mix of economic illiteracy of congressmen and corruption.
If we would make more fixed-prize contracts you could start attacking politicians who make cost-plus contracts on the basis that they cost the tax-payer a lot of money. If you try to enforce a standard of “any politician who makes cost-plus contracts that then run above budget is to blame for that”, you could shift the system to be more productive.
How would you envision the logistics of enforcing such a standard?
You write newspaper articles that blame the responsible politicians. You say that their actions resulted in a lot of wasted money and talk about how they took campaign donations from interests that profit from the government being responsible for the losses instead of private industry.
Elon Musk pushed for more space contracts to go fixed-price. In both space and defense, you could research who’s responsible for moving things in the right direction and who blocked it and resulting in wasted money.
If someone at Vox would decide they want to do something good for the world they could do it.
I don’t see how this could overcome the counter-efforts of those who currently benefit from cost-plus contracts. They after all have a lot more to lose, individually, then a society of several hundred million, where the per person costs may be a couple hundred dollars total in any given year.
Getting enough votes and maintaining voting discipline to enforce any standard at all is incredibly tough in the U.S. elections systems.
If a project goes fails and you write about how John is responsible for wasting a lot of tax-payers money because John decided to to a cost-plus contract instead a fixed-price contract, it’s hard to argue that John isn’t blameworthy if you can’t say who’s supposed to get the blame.
Avoiding opportunities to get blamed is a very strong motivator for many politicians.
What it takes, is enough reform-minded journalists who are willing to consistently talk about it for 1-2 decades.
This would work if it was the only contentious topic at stake during an election. However in reality, given recent trends, there will likely be dozens of hot button topics at stake and only a few viable candidates, and virtually all other topics carry more emotional appeal, and more motivated voting blocs, then fixed-price contracting standards.
It seems exceedingly unlikely that this issue would get enough oxygen for it to be decisive in selecting any elected candidate.