Uh, AMD drivers have most assuredly not always not just worked. They do now, and they have for something like 10 years, but before that they were a steaming pile of locked in garbage.
not to split hairs, but I think the parent is justified in saying they “always worked” if they’ve been this good for a decade.
If I was 10 years younger than I am today, my perspective would have been that it “always worked” and at some point we have to acknowledge that there has been good work done and things are quite stable in the modern day. 10y is not a small amount of time to prove it out.
>I think the US Gov probably "incentizied" Nvidias stake in Intel, and I wonder if they did here as well.
They definitely did, Intel existing is probably an issue of national security at this point, if Intel fell then there'd be the risk of some other nation's company being part of the duopoly.
> They definitely did, Intel existing is probably an issue of national security at this point, if Intel fell then there'd be the risk of some other nation's company being part of the duopoly.
Mind elaborating? Who are the players in the duopoly?
I'm not sure why Arm is in parenthesis twice, when it's a full-blown, non-American CPU designer on whose coat-tails Apple and Qualcomm have been riding.
Risc-V moved HQs to be a non-American CPU designer, but perhaps you don't find them credible (yet).
Apple and Qualcomm only use ARM ISA at this point.
And no, Apple and Qualcomm are the standard setters in ARM these days. Should they drop ARM for something else... ARM will be on the same trajectory where MIPS ended up.
RISC-V is just an ISA standard, the standard body is not a CPU designer in any shape or form.
Or they could be referring to the Wintel monopoly (Windows+Intel), or the x86 duopoly (Intel+AMD), or the FPGA duopoly (Altera=>Intel + Xilinx=>AMD)...
Global Foundries sent their EUV machine back (and paid a fat restocking fee to do it), they've stopped trying to compete at the leading edge of logic processes.
SMIC has a DUV multi-patterning 7 nm node which is already economically uncompetitive with EUV 7 nm nodes (except for PRC subsidies) and the economics of DUV only get worse further down, but at least they're trying and will certainly be the first client to use the Chinese EUV machines, whenever those come online.
Not a direct competitor, they are at a No3 slot behind Ericsson with a small global footprintmainly concentrated in NorthAmerica and some EU markets.
However most of the 5G/5G+ patents are Huawei owned and FRAND so in any case the entiti in the drivers seat is H , thas why even the whole OpenRAN project didnt get far.
Most likely like you surmiseits a geo-political hedge play.
Yep the big three plus Huawei with a bit of an edge on them with te standard essential patent , that they collaborate in a pool with.Although in the matter of mobile modems/radios Qualcomm has an edge over all the others - not so much in the backend/longhaul telco space.
Additionally if i recall most of the 6G stuff is being pushed by Huawei since most of it rests on the current 5G/5G+ work.
> I think the US Gov probably "incentizied" Nvidias stake in Intel, and I wonder if they did here as well.
If you wanted something in the x86 space it was either Intel or AMD. AMD is a direct competitor. If I was Nvidia I'd have done something about Intel. At least stop them from crashing further.
Why is the "DNS Planner" and "DNS Enactor" separate? If it was one thing, wouldn't this race condition have been much more clear to the people working on it? Is this caused by the explosion of complexity due to the over use of the microservice architecture?
> Why is the "DNS Planner" and "DNS Enactor" separate?
for a large system, it's in practice very nice to split up things like that - you have one bit of software that just reads a bunch of data and then emits a plan, and then another thing that just gets given a plan and executes it.
this is easier to test (you're just dealing with producing one data structure and consuming one data structure, the planner doesn't even try to mutate anything), it's easier to restrict permissions (one side only needs read access to the world!), it's easier to do upgrades (neither side depends on the other existing or even being in the same language), it's safer to operate (the planner is disposable, it can crash or be killed at any time with no problem except update latency), it's easier to comprehend (humans can examine the planner output which contains the entire state of the plan), it's easier to recover from weird states (you can in extremis hack the plan) etc etc. these are all things you appreciate more and more and your system gets bigger and more complicated.
> If it was one thing, wouldn't this race condition have been much more clear to the people working on it?
no
> Is this caused by the explosion of complexity due to the over use of the microservice architecture?
no
it's extremely easy to second-guess the way other people decompose their services since randoms online can't see any of the actual complexity or any of the details and so can easily suggest it would be better if it was different, without having to worry about any of the downsides of the imagined alternative solution.
Agreed, this is a common division of labor and simplifies things. It's not entirely clear in the postmortem but I speculate that the conflation of duties (i.e. the enactor also being responsible for janitor duty of stale plans) might have been a contributing factor.
I would divide these as functions inside a monolithic executable. At most, emit the plan to a file on disk as a “—whatif” optional path.
Distributed systems with files as a communication medium are much more complex than programmers think with far more failure modes than they can imagine.
Doing it inside a single binary gets rid of some of the nice observability features you get "for free" by breaking it up and could complicate things quite a bit (more code paths, flags for running it in "don't make a plan use the last plan mode", flags for "use this human generated plan mode"). Very few things are a free lunch but I've used this pattern numerous times and quite like it. I ran a system that used a MIP model to do capacity planning and separating planning from executing a plan was very useful for us.
I think the communications piece depends on what other systems you have around you to build on, its unlikely this planner/executor is completely freestanding. Some companies have large distributed filesystems with well known/tested semantics, schedulers that launch jobs when files appear, they might have ~free access to a database with strict serializability where they can store a serialized version of the plan, etc.
I mean any time a service goes down even 1/100 the size of AWS you have people crawling out of the woodworks giving armchair advice while having no domain relevant experience. It's barely even worth taking the time to respond. The people with opinions of value are already giving them internally.
> The people with opinions of value are already giving them internally.
interesting take, in light of all the brain drain that AWS has experienced over the last few years. some outside opinions might be useful - but perhaps the brain drain is so extreme that those remaining don't realize it's occurring?
Pick your battle I'd guess. Given how huge AWS is, if you have Desired state vs. reconciler, you probably have more resilient operations generally and a easier job of finding and isolating problems, the flip side of that is if you screw up your error handling, you get this. That aside, it seems strange to me they didn't account for the fact that a stale plan could get picked up over a new one, so maybe I misunderstand the incident/architecture.
This was my thought also. The first sentences of the RCA screamed “race condition” without even having to mention the phrase.
The two DNS components comprise a monolith: neither is useful without the other and there is one arrow on the design coupling them together.
If they were a single component then none of this would have happened.
Also, version checks? Really?
Why not compare the current state against the desired state and take the necessary actions to bring them inline?
Last but not least, deleting old config files so aggressively is a “penny wise pound foolish” design. I would keep these forever or at least a month! Certainly much, much longer than any possible time taken through the sequence of provisioning steps.
I also really like dev containers in vscode. Wondering if you've looked into using something like fedora silver blue and it's "toolbox" system to do the dev container business at that level.
I agree from a company perspective using dev containers gets all devs on the same page
Cloudflares docs are written so hard for web dev. Can you host a monolith app that isn't serving http traffic on cloudflare tech like containers? Like can you spawn a container and have it handle tcp or udp connections until you manually shut it down? The container docs say they auto shutdown after not receiving requests...
It's a really good platform for Typescript microservices which scale-to-zero (up to very high theoretical limits), but it wouldn't be a platform you'd migrate a monolith PHP app to (for example).
Rewatching the video it strikes me that there is no talk of how the compute will be paid for. Altman is saying there is so much demand for compute but a lot of that is because he's setting the price for users to zero in spite of the service costing a lot to provide.
I guess they figure once they have the users they'll monetize somehow but that bit's kind of iffy.
I agree, though it's not out of the question that there could one day be a single country whose economy relies entirely on data centers, similarly to petrostates. Maybe we'll call them datastates, or compustates.
Whenever a company is aligning their earnings and always beating them, you should ask questions. If you've ever read the book "Smartest guys in the Room" that's the biggest tell.
It's a market where line has to go up, so bad behaviour is demanded.
Giving OpenAI $30 billion, so that they can spend $30 billion on GPUs, so that you can now report your own $30 billion as revenue is market manipulation and an attempt to get others to invest in this "$300 billion" market, so OpenAI can pay their bonuses and can give Nvidia their 40%.
LLMs might be really good at converting your USB debugging script from Python to Go, and it might do a passable enough translation that you don't need to use Google translate but that's not a $300 billion market.
Amazon cooked their books into the negative for years and only recently went positive. All nvidia has to do it pay out the gov't ( like they already are) and things are fine.
There is a lot of suspicious behavior, and if the top analyst at Morgan Stanley is saying it - their chief investment officer - then I think there is at least some smoke. The suspicious behavior could be considered the flames.
> National post is a respectable publication more akin to the Wall Street Journal.
Absolutely not, their American owner has cut so much into that newspaper that whatever reputation they had a decade ago is entirely gone. They used to be somewhat of a right leaning equivalent to the Globe and Mail, but this is no longer an equivalent, what comes out of that publication is now extremely poor quality, often entirely false.
At best you could compare them to fox news, but they don't compare favourably even to that. They are admittedly maybe not quite as bad as Breitbart but it's close enough in my opinion.
The globe falls on the conservative side of things. The star is currently in a weird middle ground where it was somewhat left but it is being moved somewhat right by its new owners. CBC is reasonably neutral.
Highly partisan news is generally not respected for good reason though.
The globe used to fall on the conservative side of things, now honestly the Toronto star is more to the right than the globe. Both got bought by new owners in the last decade and it shows.
Interesting choice of words - hasn't it been shown, repeatedly, time and time again, that conservative-leaning publications are generally not respectable? You want a neutral publication, not one with a lean.
The Atlantic is a highly respected publication and an argument can be made that they are conservative in that they promote a kinda of neo con foreign policy through David Frum et al's work
Are there any, anywhere in the developed world? Le Figaro in France might count, but pracically every explicitly socially conservative media I can think of has had at least some suspicious events that have lowered or destroyed its reputation.
Be it Fox News lying and then claiming they're entertainment, not news, so they're allowed to lie; The Telegraph giving a soapbox to various climate skeptics and COVID deniers and having to retract them, or publishing flat out lies by folks like Boris Johnson; the Sun needs no explanation; Sky News Australia (same owner as Fox News) having a "Misinformation and conspiracy theories" section on their Wikipedia; etc.
I know how to use the terminal to enforce deep sleep on laptops, but thats about all I do setup wise.