But a given launch system is deeply intertwined with designing the mission. The size, the weight and the mass distribution.
Moving parts are a nightmare for reliability. Every moving part is something that can cease up and go wrong. It's a motor that can fail and gears that can get jammed.
JWST has two key components that involve a lot of moving parts:
1. The mirror. Hubble was smaller enough to be deployed fully constructed in the Space Shuttle so didn't have to deploy in space. JWST's mirror is 3-4x larger an there's no current launch vehicle that could launch it fully assembled. That's why you have all the beryllium hex mirrors that had to deploy the mirror once in orbit. These motors, actuators, assemblies, etc need to be incredibly precise and reliable; and
2. The heat shield. JWST has to be incredibly cold to operate (5K IIRC). The only way to get rid of heat in space is to radiate it away. The heat shield separates JWST from the Sun, the Earth and Moon (each of which reflect enough light to interfere with operations). The shield is several layers and it's large, like tennis-court sized. Obviously this too had to be deployed in space.
There were like 10 technologies for JWST that had to be invented to make the mission possible. That's less than ideal. It adds to the cost, the complexity and the timelines.
In hindsight it probably would've been worth having a stepping stone between Hubble and JWST that proved some of these technologies in a cheaper and less risky way, probably with a smaller mirror. But here we are.
But it seems patently obvious. After all, 10 technologies don't have to be reinvented. No research would need to be redone. No test rigs would need to be redesigned and duplicated. And on and on.
The most bizarre argument against building a duplicate was there wouldn't be anything extra worth looking at. Yet I watched the NOVA episode on the scope last night, and everywhere they look where we thought there was "nothing" turns out to be crammed with 10,000 galaxies and stars.
Consider fabbing chips. Fabs spend an awful lot of time verifying the produced chips. I've seen estimates that verification cost exceeds fabrication cost.
How much of the development cost of JWST 2.0 would be spent on verification? I honestly don't know but I would guess it's high.
Another factor: part of making a production process is predicated on how many you'd build. We had a process for making Saturn-V rockets based on the number of Apollo missions we planned for. If you then up and decide you need 500 Saturn-Vs you might have to go through a whole new process for something that will scale that high.
Yet another factor: the launch vehicle. If you decide to do JWST 2.0 in 10 years, what launch vehicle will you use? The same one might not exist so the new mission will have to be designed for what is available.
And another: materials science changes. We don't make the same materials that we did 50+ years ago for good reasons but those materials are a key part of the design of something like Saturn-V.
So I imagine JWST 2.0 would be cheaper but 90% cheaper? I'm not convinced.
L1: between the Sun and the Earth. The Sun reflecting off the Earth would interfere with instruments.
L3: This is suitable for positioning but impractical because of distance from a communication and perspective.
L4 & L5: these are unsuitable because they're stable. Distance is an issue too. Why are stable Lagrange points unsuitable? Because things tend to collect there. Over the eons a whole bunch of crap has built up there and will continue to do so. That's not what you want when dealing with sensitive instuments.
Now you could say that you could orbit L4 or L5 like JWST orbits L2 and that's true but you still expend energy for stationkeeping. At L2 you spend energy to stay around the L2 point. At L4 and L5 you spend energy to avoid falling into the Lagrange point.
According to [1][2] it's more expensive (in delta-V terms) to get to L4/L5 than L2 but not hugely so.
[1]: https://space.stackexchange.com/questions/27010/how-does-the...
[2]: https://space.stackexchange.com/questions/57463/what-is-the-...
That still leaves the L2s of Venus, Mars, Jupiter etc.
And you can probably fit a few more into Earth L2.
It has an orbit around L2 of roughly 0.8 million km. L2 is about 1.5 million km from Earth. The moon is only 0.36 million km from Earth.
* L1 is between the Earth and Sun, so orienting the sun-shield toward the sun would result in the cold-side being heated by reflected light from the earth. It would also have access to the same patch of sky as at L2 for any given time of year, so there wouldn't be an advantage to having it at L1 as opposed to having a second telescope at L2.
* For L4 and L5, there's a ± 60 degree orbital offset relative to Earth. That gives access to a different part of the sky at any given time, but over the course of 6 months, both locations could access any part of the sky. There's a potential for time-correlation studies, where an observation at L2 and L4 are made at the same time, with the ~150 million km between them serving either as the base of a triangle for parallax measurements or as a known distance for measuring speed of travel between the
That said, I'm not sure how useful those would be. The parallax measurements could be done by ground-based telescopes simply by waiting a few months between measurements, and the speed of travel measurements would primarily provide a measurement of the index of refraction inside the heliosphere.
Overall, I'd guess the biggest scientific bang for the buck would be to load up more telescopes at L2. The proposals for the first 12 months of observations totaled about 48 months of observation time [0], so most of the proposals were rejected. The biggest advancement I can see would be to simply have more telescope time overall.
[0] https://www.stsci.edu/contents/news/jwst/2020/jwst-cycle-1-g...
[1]: https://www.esa.int/Science_Exploration/Space_Science/Hersch...
That's what I supposed.
Even L2 is big. I'm sure you could 10s or 100s of kilometres either side of "L2" and still have the same benefits as Webb has. In fact it you have enough objects there, and cooperate, they could effectively orbit each other.
Not a concern.
First: The Sun/Earth L2 is a big place. JWST's orbit around L2 is substantially larger than the Moon's orbit around Earth, and its orbit moves around in multiple dimensions -- it isn't constrained to a single dimension, like a geosynchronous orbit around Earth is. Even if there were debris in the vicinity, the odds of hitting it would be negligible.
Second: Orbits around L2 are inherently unstable -- it's a "saddle point", not an attractor. JWST has to maneuver a bit to keep itself in orbit, so any debris that ended up around L2 would naturally drift away.
Also, the entire Apollo infrastructure was actually designed for many more launches than occurred, that's why some of that infrastructure is still in use to this day. It's been a few years since I read a biography of Von Braun but I remember him making design decisions for Apollo that anticipated many more launches and manned missions to Mars and Venus in the 1970s with a manned Mars base by the early 80s.
The satellites which were donated to NASA by the NRO where likely leftover from the abortive Future Imagery Architecture program which was unrelated to Hubble.
Source: just my perspective after 20 years working on space systems.
Case in point: When there are 3 people on a planet that know how to create a specific sensor, and they happen to be NASA civil servants, JPL employees, Professors in Academia, or from a boutique non-profit like Aerospace or Mitre...in many cases the govt will decide to push the fabrication to commercial industry to create it, instead of the elite teams that created the most recent groundbreaking sensor.
This causes huge cost overruns as the commercial teams fumble around [1] and in some cases, even get paid despite dropping a satellite [2]
One of the funny fixes to this for NPP, was assigning a government "hall monitor" experienced in safe satellite testing to oversee test operations, and gig them for every infraction of safety protocols.
[1] https://www.defensedaily.com/npoess-has-more-than-1-billion-...
[2] https://spacenews.com/another-factory-mishap-damages-noaa-n-...
But of course. I meant before that happened, i.e. planning to make some siblings.
basically after the manufacturing, calibration, tuning, assembly, etc. techniques have been perfected to the required degree
plus, it would make sense to try to repeat this a few times always trying to 80-20 it. (I mean intentionally trying to "half-ass" the next iteration. spend less on it than on the previous one, and try to get the best out of that budget - with the same tech of course.)
the big problem is that ... it doesn't make much sense to do this. JWST delivers the data for the science it was intended and ... for the new questions that we'll raise we'll probably need some even more specialized observatory.
Just for fun, buy two IKEA flatpacks for, say, bookshelves. Build one, then the other. I bet you spend half the time building the second, even though you're doing exactly the same thing, by hand.
Reminds me of when in March 2020 I was also told I didn't know what I was talking about when I pointed out how a Covid vaccine could be made available in 6 months rather than the conventional 18 minimum times. The vaccines were available in 7 months.
The airplane I worked on the design of, the Boeing 757, would have cost at least 100 times as much per-plane if it was a one-off design.
It's not clear to me that designing something that needs to be assembled successfully once vs twice vs 10 times doesn't introduce some more cost tradeoffs in itself. Just around tolerances and yield even, how many pieces are required to be made to give one complete set of parts that all work together vs 10 complete sets that all work together. I could see some extra pain and assembly from having to get more permutations of part variances all working together and validated. A while back, at least, a lot of the testing processes for space components were VERY manual so you'd either need to invest more for automation or accept linear costs on all the extra production/verification.
This is partly quibbling over trivial details - I would imagine the savings is closer to 50-75% than 90% - but at a certain point those differences in tens-of-percentage points are lots of billions of dollars.
Here's the real kicker to me: why spend even an extra 10% up-front before even deploying one of them? Why not spend the marginal costs on building an additional unit only if the first fails? Then you can learn from the failure too? And if it works the first time, those billions can be spent elsewhere instead.
You're right that the 757 was meant to be built in bulk, and hence one has to pay attention to tolerances so the parts are interchangeable.
Conventional vaccines are sterilizing, have good safety records, and protection in years, or decades.
This "vaccine" is none of above.
These MRNA "vaccines" required a change in the definition of a vaccine, and have atrocious safety records.
The new Novavax vaccine [1], which is a conventional vaccine, may prove to have much longer effectiveness period, and not require boosters every 3 months. It also took much longer to develop.
[1] https://www.novavax.com/science-technology/recombinant-prote...
It's great that newer, vaccines are being developed. But in the meantime, the Novavax vaccine is not available. I'll take the one that is available. I'm not really interested in dying waiting for the perfect vaccine.
Many vaccines do not provide sterilizing immunity. Just to give one random example: flu vaccines do not provide sterilizing immunity.
> have good safety records
The various SARS-CoV-2 vaccines have very good safety records. They've been administered to billions of people, and the rates of serious side-effects are on the order on one per several hundred thousand - orders of magnitude lower than the rate of serious complications from infection.
> The new Novavax vaccine, which is a conventional vaccine
I don't know how you're defining "conventional" here, but Novavax also uses a new vaccine platform.
I know 2 people with severe adverse events from the vaxx.
Considering how many vaccinations I have received my life due to military service, that is an order of magnitude worse than any vaccination, ever.
Those are the ignorant ones.
It isn't quite a "Those who know don't speak and those who speak don't know" situation, but this is just blindingly obvious to anyone who has done any R&D to manufacturing and risk assessment. Heck, even just having a duplicate on the ground to debug could save the mission (since there isn't one, let's hope it doesn't come to that).
On one-off projects of any size, the design, prototyping, testing, & refining the design just overwhelm the cost of fabricating the final parts — and they do it by orders of magnitude. Just for the carbon fiber parts I work in daily, the initial R&D test program for a sizeable (scale of 1 m^2) part might cost $25K, the first mold $12K, the first part sells for $3K, the second part $2K and the fifth part 1.7K. To get to volume production in the 100s, there's probably another dev program & set of molds, and the 300th part out the door might sell at $1200.
Modern cosmology assumes-so, but it need not be true.
Homogeneity has been very well tested out to a redshift of a few by surveys like the Sloan Digital Sky Survey and the Dark Energy Survey, each of which have observed thousands of square degrees of sky (there are about 41k square degrees in total). And as another commenter notes, the CMB (at redshift 1100) is homogeneous (to one part in 10k).
What if 10 were built, each with only a 10% chance of success? What would that have cost? After all, it doesn't need the expense of being man-rated.
There’s physically not enough infrastructure (clean rooms, testing facilities, vacuum chambers, etc…) nor skilled manpower available to NASA to be able to build more than one in parallel, so that would add to the cost significantly.
Nothing?
Of course, the sensible thing to do is to build the test equipment, procedures, train the testers. Then test #1. Then test #2. Then test #3. That's what everybody does.
The cost of the Saturn V rocket was amortized over many launch vehicles, despite a host of new technologies that had to be developed for it, despite much of it being hand-built.
All that adds to the cost of the first mirror, and adds $0 to the second.
I don't think these NASA/JPL guys are quite as dim as you seem to suggest (but you have more info) and they may have done a little cost/benefit analysis before embarking on the project.
BTW, I am not suggesting they are dim. But one can be the finest physicist in the world and not be familiar with cost accounting and manufacturing.
Successful launch and arrival at the observation point? The 6 month setup period before observations can be made?
I contend that success here is a full mission that yields science data over decades of observation. If we cut down the acceptable error rate so we can launch early, how does that impact longevity?
2x — failed to deploy, mission lost
4x — early operation error, 6mo lifetime
3x — lower performance, 3 year lifetime
1x — mission successful, 10 year lifetime
…do we get a better deal building them with larger faults, at a lower cost?
A lot of expense is in developing technologies, assembly tooling, and test rigs — all of which is easier if we don’t need them to be as assured because only 1 in 10 needs to succeed.
No idea. Do you? Genuinely interested, particularly regarding methodology. Not intending at all to be insulting.
My dad told me that Pratt&Whitney made the most reliable aircraft engines by putting them on a test stand and running them at full power until they broke. The engineers would figure out why it failed, redesign the part, stick that on the engine, and continued running them at full power.
It's straightforward, inexpensive, and it works.
Then there's all sorts of complex failures that aren't addressed by single engines: " Prior to this crash, the probability of a simultaneous failure of all three hydraulic systems was considered as low as one in a billion. However, statistical models did not account for the position of the number-two engine, mounted at the tail close to hydraulic lines, nor the results of fragments released in many directions. Since then, aircraft engine designs have focused on keeping shrapnel from puncturing the cowling or ductwork, increasingly utilizing high-strength composite materials to achieve penetration resistance while keeping the weight low"
(I'm not disrespecting P&W- I'm sure they have more tests than just "100% power until it breaks, fix and repeat")
Keep in mind that my former job was designing gearbox parts for the Boeing 757. This included doing the math, and devising a test plan. I've spent a lot of time on "what could go wrong" scenarios in the real world.
Also, when I prepare slides for a coding presentation, the implementation code for a concept gets trimmed way down to what will fit on a slide.
That doesn't really seem to be in NASAs DNA, to do smaller incremental improvements. Instead, they really upgrade big, in chunks with long time in-between. I don't know why they do it, but in their webcasts it been mentioned a couple of times related to James Webb.
Also remember that a non-trivial amount of Americans believe the moon landing was faked, or that the earth is flat and NASA is part of some conspiracy. Or that a certain political party has spent literally three decades saying "the nerds at NASA are lying, there is no global warming" and you might start to understand the complete lack of political will to give NASA the funding to do even literally the basics.
There’s typically only a chance to build a flagship every ten years or so.
Maybe the main reflectors could be like that too. One or more packages that are just the reflector, maybe not even all of it but just sections.
I think the adjustable array of smaller sections is simply necessary to get and maintain the focus anyway, so the adjustability is not an expensive extra.
Hubble actually went up with a botched main reflector and couldn't be corrected directly. The replacement secondary optics was only a terrible patch, and really only even possible because the scope is within human reachable space to do that level of heart surgury in-situ like that.
So if the reflector has to be adjustable anyway just to work at all, then it's not a crazy extra fundamentally, and only a bit extra to have the pieces be able to be shipped individually and self-assemble on site. The inherent adjustability which was required anyway, allows for the inherent inaccuracy of connecting two parts vs one solid part.
With 2 or 3 or more smaller component missions, each project becomes not only less critical and more tolerable of a failure, but also less complex and less likely to suffer a failure.
And it seems like this project does have at least 2 or 3 natural major building block boundaries. The shield vs everything else is one for sure. That could be turned into a platform, maybe for more than one instrument added or replaced over time, and could maybe supply power as well.
Then maybe also the main reflector vs the actual instruments. And the main reflector itself could be in simpler discrete sections instead of having to fold and unfold.
Tennis court sized shield? Take it up like a long and precious Persian mat and let the humans rig the thing in orbit.
From there, attach fuels tanks and you are good to go for your insertion burns.
The sun shield is held up as this big challenge but I don’t think they do a good job explaining why it was so difficult.
There’s a popular YouTube whose father was a metrologist working on the sun shield. https://youtu.be/Pu97IiO_yDI
Is it impossible to build to a lower level of reliability then have humans available to fix/adjust things in orbit?
Then once it's fully assembled in place, move it to its final location?
"Aligning the primary mirror segments as though they are a single large mirror means each mirror is aligned to 1/10,000th the thickness of a human hair. This alignment has to be done at 50 degrees above absolute zero! [0]
[0]https://www.nasa.gov/topics/technology/features/webb-actuato...
1. Where doe the telescope deploy? Ideally in LEO so you have a chance fo fix things but then that complicated the launch. You have to park in orbit and restart th eengine at a later point. AFAIK JWST didn't enter a parking orbit so would've required additional fuel. That might not have been possible;
2. What if the launcher itself fails? Do you want to have the capacity to launch another rocket, have it rock with JWST and then go on? If so, that's a whole new level of complexity and a set of problesm you have to solve as well as things that can go wrong;
3. If you look at the flight plan there are a bunch of turns. How would the G-forces affect, say, the deployed heat shield?
4. JWST actually had to rotate during launch to point the instruments away from the Sun so as not to destroy them. This already added complexity. Doing this with the deployed spacecraft would probably only further complicate this.
There was a tremendous amount of r&d expense for James Webb. However I think that sibling comments are underestimating how much can be reduced by just expecting that the first few attempts would fail. Take the sunshield, where Smarter Every Day did a video showing the extent to which everyone worked to ensure the exact shape was perfect: https://www.youtube.com/watch?v=Pu97IiO_yDI
A lot of this difficulty was because you couldn't just put it up there and see if it works. If an extensive test costs $100 million, but a launch costs $177 million, you choose the extensive test every time. I think overall a program which made dozens of James Webbs, launching a couple times a year would likely have been cheaper with better results.
On the other hand, there's a big problem: Failure, even within the expected threshold of failure, looks really bad in the realm of public opinion. There's just the practical problem of a NASA director having to stand in front of congress and justify why the telescope program that has launched 6 failed satellites over the last 3 years should still get funding.
but this is still just an interested observer's speculation.
**
1. The telescope and instruments are designed for a one-time special use, shared by no one else (pretty much). That means many if not all the requirements are being discovered as they go, and it is not a very predictable process in terms of the risks and unknowns. Many delays happen/happened because people believed the requirements were set, and started building things, only to have them change later, leading to wasted time and resources. But that's how something in research phase goes.
2. The size, materials, instruments, etc were pretty much unprecedented, aside from some very low level legacy components. Everything had to be designed for the first time. This often leads to many unexpected cost overruns.
3. It was a huge project, which is always in tension with something that is R&D / being discovered as you go. It takes a long time for many different and scattered teams to be able to communicate their requirements and capabilities and schedules to each other when they are so distributed. There really is something to the idea that 6 people in one room can do things that 100 people in multiple rooms cannot, but by the sheer size required, and the fragmentation of expertise to do the job, it had accompanying schedule and risk problems.
4. Because the telescope is a one-off and so valuable as a project, it had to be risk-free or risk-minimal. When something costs so much, it has to cost even more to protect it against mistakes. That means that you can't cheap out and risk certain things, leading to it taking more time and resources to get it right. It is the opposite of the Mars program "fast and cheap and fail quickly". It has advantages and disadvantages.
5. Also a sort of less tangible factor is that these projects, even when delayed, have to keep a certain group of people employed to maintain continuity of knowledge and technical expertise. If you get delayed, you cannot just cut people from the program, you will lose them to other projects and further delay progress. So every year of delay incurs you a very high cost.
**
The costs were not the cost of launch. That was a relatively small part of the project. Ariane, etc. are known factors at this point. It was the fact that everything about the telescope was new, will never be reproduced again, and had to be gotten right the first time.
I am probably oversimplifying some things that happened during the process, and other related factors, but that's my opinion.
The second part, 'changing requirements' has been the reality of many NASA and DoD projects for decades.
The "delays" happen because the vendors make gobs of money working this way.
I would also say that it is a product of the federal government in recent decades gradually losing its ability to keep people on staff (or pay them enough) to build the knowledge about how to run/build such projects themselves. And cost-effectively.
If you recall the earlier days of aerospace, aircraft, etc., technical experts in the military would basically tell Lockheed, whoever, exactly what they wanted to design, or would be equally qualified to set out the specs and be deep in the design.
Over the last decades, that capability (I believe) has largely left the government/public institutions. We have essentially outsourced the design and building of aircraft, spacecraft to private contractors, and when that happens it naturally costs more to do, to pay them to do that job (and take on the risk of doing it). After all, they are profit-seeking enterprises, while if that expertise had been kept in government, it would not be.
If you take it by examples, the era of WW2/shortly-post-WW2 military aircraft was when Air Force/Navy/Army aircraft engineers helped design planes that contractors would get marching orders to go build (of course with their input). Nowadays, we're in the era when Pentagon procurement office tells LM / Boeing to go design us a plane to these outcomes, which are cobbled together from 4 different branches of the military and uncoordinated generals' wish lists.
And we're surprised that a fighter program ends up costing $2T.
Edit: I googled it. Here's the first paragraph of the first link from DDG:
> The James Webb Space Telescope (JWST) is expected to cost NASA $9.7 billion over 24 years. Of that amount, $8.8 billion was spent on spacecraft development between 2003 and 2021; $861 million is planned to support five years of operations. Adjusted for inflation to 2020 dollars, the lifetime cost to NASA will be approximately $10.8 billion.
So "spacecraft development" comprised about 90% of the total budget. Of the remainder, 96% is allocated to "five years of operations". So we have about $39 million left over, which I assume covers the launch and everything around it. If you want to try optimizing that, go ahead, but it is nearly a rounding error (0.4% of budget) in the grand scheme of things.
There is an argument to be made that a few million dollars saved is valuable. I would propose a counterargument, that the amount of friction induced by working with private contractors (especially contentious and manipulative ones like those who run SpaceX) would add to some of the development costs, so the savings may not be as big as you think from a naive comparison.
It's like the silly analogy I hear now and then that sending a probe into the solar system is like throwing a baseball from LA to NYC and hitting a small target there. It isn't like that at all. The probes undergo course corrections as necessary.
How much? How much did it cost for the actual JWST? Do you have any evidence of this great cost?
The glorification of Musk's ego is endless. Can JWST see the farthest reaches of it? Is it expanding or contracting?
If that isn't convincing enough, the budget plan including all folding mechanisms etc before testing was $6.5B, so the $3.5B extra was spent entirely on testing and repairs. Thus, if a folding mechanism was not needed, they could have saved at least that much on testing and fixing said mechanism.
Might be a good idea to take off your blinders before complaining about someone being more successful than you. Since your jealousy is blinding you so much, it would have been just as convenient for costs if any other rocket had a wider payload fairing than Ariane 5.
[0] https://www.space.com/james-webb-space-telescope-deployment-...
Nope! It doesn't cost 8 billion to unfold a mirror. Seems like a rather straightforward mechanical problem to me. And your baseless theories and mine are worth nothing without evidence. That's the reason we built the JWST - we need evidence. Baseless nonsense is the stuff of cults of personality, not knowledge. We need much more of the latter and much less (i.e., none) of the former in our society.
The mirror segment deployment technology is surely a significant percentage of the cost.
Is there a more common or obvious trap of reasoning? If 'surely' worked, we wouldn't have needed the Enlightenment, science, or the JWST to learn about the universe.
IIRC (much stronger evidence than 'surely', but still lacking), it's in fact not significant to the cost.
Is there a more common or obvious trap of reasoning? If 'IIRC' worked, we wouldn't have needed the Enlightenment, science, or the JWST to learn about the universe.
I know someone who was one of the test engineers on SDO (solar dynamics observatory) and there were all manner of concerns about this, that, and the other thing that were really hard to say were almost absolutely certainly "just fine."
As a simple analogy: there's a difference between going to 3-for-1 Suite Warehouse and going to Saville Row and getting something bespoke.
It's just there are no COTS satellites that do what JWST do, so if you want to do cutting edge science you're paying for Saville Row.
I don't think you know anything about the space industry. SpaceX is famous for being wildly easier to work with than any other launch provider (including semi-national providers like Arianespace).
This doesn't really carry-over to large projects like JWST yet though.
Not to mention the spacecraft and all the deployable systems must withstand intense G forces to achieve escape velocity.
The design also requires complete verification. Each component must be created and a test bench then has to be engineered to ensure viability after launch. It is an immense undertaking to develop experimental equipment (ultra-high vacuum, pulsed powered laser physics background), to then add the additional expectations of space launch and zero opportunity for corrective intervention means the standards are exacting.
When we get into the government allocating funds it gets tricky. Cheaper launch costs could just mean that the savings go into the telescope rather than risk management. It's hard to get funds for a backup telescope, even if it's cheaper overall.
In my head (and I totally have nothing to do with NASA/JWT), it’s the reliability requirements.
I have this discussion with clients a lot, I had it just last week. They’ll come to us with a project with some vague requirements and I’ll ask “what are your reliability requirements here?” and very often they claim “this is business critical, it needs to be 100% reliable”. So I go into my spiel about how 100% is not an achievable goal, and we need to be realistic about whether their actual requirement if 99.9999% or 99% or somewhere in between. For web/app-backend/*aaS type projects I’ll point out I can provide 99% SLA hosting for around $100/month, but that each extra “nine” costs ten times more. If you want guarantee less than an hour of downtime per month, it’ll cost you $1000/month to host, if you want less that 5 minutes a month of downtime I’ll charge 10,000. If you need less than 30seconds a month of downtime we’re talking an on-call team of 5 engineers on 24x7 rosters and that’ll cost you $100,000 a month.
If I were launching a one off custom built billion dollar telescope to an orbit a million miles away with practically zero chance of ever fixing anything that might need “hands on” or replacement, I’d be looking very carefully at the costs for 6, 7, or 8 nines of reliability. And I’d choose to spend a _lot_ more than most people would consider “reasonable”.
There’s a difference between monthly web hosting costs and the JWT reliability, but instead of 10 or 100 million a month over the life of the project, they need to spend a similarly astronomical budget on the architecture, design, manufacturing, and testing all up front.
Can you imagine designing a web hosting platform where the requirement was to run for 5+ years with zero downtime and zero human intervention? U can’t “more hardware” your way out of reality there, and you can’t ignore the risk of your cloud hosting going down (or under). So you now need to think along the lines of 8 nines of HA platform with exception handling and self healing for every possible scenario, in a multi cloud configuration in case Amazon goes into liquidation, that is resilient to _everything_ from as yet undiscovered OS and hardware vulnerabilities to DNS and SSL cert outages and BGP hijacks.
(That’s turning into a fascinating thought experiment for me now. I wonder how I’d plan things if someone said “I’ll pay to 1 billion dollars if you build me a (non trivial) backend hosted platform/application, that has less than 1 second of downtime in 5 years, and the _only_ maintenance you can do is reboots or software updates over a 300 baud dialup modem, no platform/hardware configuration changes allowed..)