The NixOS Basis’s Name to Motion: S3 Prices Require Neighborhood Assist – Bulletins
I simply wish to observe a number of issues concerning that.
I very a lot agree that this information is extraordinarily useful, and shedding it could be a horrible blow for analysis communities and even understanding our personal ecosystems.
That’s why I imagine, it doesn’t matter what occurs: new retailer paths must be shipped to one thing that doesn’t value egress charges if potential, extra on that later.
Onto the subsequent factor:
The $32K migration payment, is, I assume, the egress thingie that AWS cost everybody for attempting to stop their platform.
A number of issues to unpack:
(1) Have we reached out to AWS concerning this matter whereas pleading our case as a non-trivial open supply mission (which they benefited from, I’m sure, not directly)?
(2) Have we reached out to a monetary cloud consultants similar to https://www.duckbillgroup.com/ that are very lively on social medias and I normally imagine good to open supply tasks for assist?
(3) Egress charges are a identified tactic to vendor lock-in folks right into a platform:
AWS are a really well-known offender on that.
Fortunately, R2 Object Storage has ZERO egress payment for now.
I believe this example is very altering with respect to laws:
https://twitter.com/platombe/status/1656634228979781633 (French tweet on a legislation attacking the egress charges state of affairs from the cloud suppliers by the French authorities).
Sécuriser et réguler l’espace numérique (artwork. 7. §. 6)
« III. – Il est interdit à tout fournisseur de providers d’informatique en nuage de facturer, dans le cadre des contrats qu’il conclut avec une personne exerçant des activités de manufacturing, de distribution ou de providers, des frais au titre du transfert de données vers les infrastructures de cette personne ou vers celles mises à disposition, directement ou indirectement, par un autre fournisseur, à l’exception des frais de migration liés au changement de fournisseur.
“III. – It’s prohibited for any supplier of cloud computing providers to cost, inside the framework of the contracts it concludes with an individual finishing up manufacturing, distribution or service actions, charges for the switch of information to infrastructure of this particular person or to these made obtainable, immediately or not directly, by one other provider, apart from migration prices associated to the change of provider.
— Google Translate, 2023
I imagine that this might apply sufficient stress in order that AWS might yield this silly quantity of charges in our particular case, no less than, this could be a gesture of goodwill given the skinny ice they’re on.
– In fact, nothing is magic, and it could possibly be as effectively that each one these makes an attempt would fail. I simply wish to ensure that we try every part and that our failed makes an attempt can even function examples for additional leveling the taking part in area for these migrations, which we’re undoubtedly not the one ones to do.
On long-term solutions
The breakdown shows:
- 107 TiB in standard storage
- 318 TiB in infrequent access storage
for a total of 425 TiB.
Current technology gives us at reasonable prices: 30.72TB SSD (e.g. PM1643a) and 22TB HDD.
Assuming we store 150TB of “hot storage” with a fast medium (SSD): 10 professional disks of 15.36TB. (no redundancy is assumed here.) for a cost of ~11K EUR.
Assuming we store 500TB of “cold storage” with a slow medium (HDD): 23 professional disks of 22TB. (no redundancy is assumed here.) for a cost of ~13.8K EUR.
What did we miss in this small computation?
- Geodistribution? We don’t seem to care because we have (for now?) Fastly in front of it which will geodistribute the cache across the world? (Also, we have the back’n’forth situation latency between the US and EU for some current servers of the infrastructure.)
- Internet? I can speak for France easily and having 100Gbps or even 400Gbps and being plugged directly in the adequate IXP is not really complicated, I can even email get a proper quote on what it would cost. Last time I checked, having something like 10Gbps commit (it means that you get 95th percentile 10Gbps guaranteed) was around 1-2K EUR per month.
- Disk replacement? Yes! I didn’t conduct the long term analysis and I wanted to post this early. This should be computed and included in the price by using Backblaze B2 statistics for example (Backblaze Hard Drive Stats).
- Human value? Sure! This all depends on below the idea we can have sufficient individuals to deal with such a factor and function it appropriately. Is that this a good assumption? Let’s see under.
- Sturdiness? Sure! Amazon S3 offers an insane sturdiness metric, which is cool and neat: although, do we’d like it? (99.999999999% of sturdiness FWIW, sure, it has this a lot 9s. However solely 99.99% availability) — although, we will push our metric additional by piling up on redundancy and a number of tiered options as we want.
- You say it: I might have forgotten one thing on this.
This small train is essential IMHO to resolve correctly for the subsequent resolution following the rule “by no means let an excellent disaster go to waste”.
NixOS’s cache has particular properties that aren’t shared by all customers of Amazon S3 and we don’t want to remain captive in a cloud ecosystem if we don’t must.
Whereas I don’t assume it’s life like besides if sufficient folks step in, and it finally turns into life like to offer a selfmade resolution for this (be it via one thing like a Ceph filesystem or Storage object storage or MinIO object storage or ). I really feel like this has proven the necessity for our group to step in for these issues, as we undoubtedly have the experience in-house and the dimensions for it.
My proposal would most likely say: let’s attempt to experiment on these issues I stated earlier than. For the human value, as that is an experimentation, it mustn’t have detrimental impacts. Another is to contemplate scaling the infrastructure crew adequately and provides them extra energy concerning this. It’s additionally a novel place to do one thing concerning the signing key of the shop (see newest RFC on the topic for extra info).
Lastly, whereas I’m not tremendous keen on this resolution, have we thought-about college mirrors like many different distributions do it and have we thought-about working with them in the direction of that?
Anyway, I hope that my messy “chime-in” can present new leads on what we will do concerning this example. I’d discover extraordinarily regrettable:
(1) we lose the historic information as a result of AWS egress charges are huge, have we thought-about utilizing AWS Snowmobile? Or related “bodily switch” options to ship them out?
(2) we might waste the cash of the Basis on a non-stable state of affairs which might value us extra on the long term as a result of we simply get milked by cloud suppliers who’ve insane margins as a result of IMHO the tradeoff for utilizing them is talent/competency/time/human value — nixpkgs is stuffed with that! Let’s make it shine.