Because they give a lot away and post blog posts attacking AWS folks think they are cheap. But a lot of their free stuff may be in the marketing budget. The issue here is:
1) Their paid offerings for their core product, CDN, does not have a public pricing guide. That's a BAD sign if they claim to be competing on price. So we can pretty safely assume their paid service bandwidth charges are NOT the 80X cheaper than amazon despite their claims here.
2) They hype their bandwidth alliance. I tried following this up on Azure, it's basically all call a sales person / maybe you'll get a deal type stuff. Smoke and mirrors again.
3) If you actually try to use their CDN to serve something high bandwidth (games / software updates / video streaming) that is going to violate some type of clause which says that the free CDN is not REALLY unlimited.
4) Financially they spend a mind boggling amount on sales costs. Between sales costs and cost of sales I think they are negative relative to revenue.
BTW - If you try any of the other "free" and "unlimited" providers (for $5/month etc) you will also find out once you try to move some real data volumes (video etc) - it's all a lie over and over.
Same things with geo issues. Cloudflare claims to have "one" price globally. These are almost alwyas false claims. It's easy to test - setup a stream into a high cost area, stream a bunch of video there - you'll usually get a call over some tricky TOS issue they pull out.
In short, if you want pay and forget, there is a reason the big cloud players have a business still.
Enterprise customer here. throw away account, since not sure how much I should be able to divulge...
1: we are paying (ballpark) 5-10k for our account. We have 5-10 domains on the enterprise plan, and we have roughly 100m requests per month plus about 70TB bandwidth, IIRC... Mind you, we have other sites that are on Pro ($20) or business ($200) plans and they are coming in at around a million requests a day (only about 2-3Gb a day) and no one complains.. could be the 5-10k we spend a month with them makes them think twice... Most of our cost is for the 24/7 tech support, regular meetings, etc. extra domains are a somewhat fixed cost (little over double a business plan domain) and that gives us extra features.
2: for Azure and BW Alliance, we were told you need to use a storage account and use "Internet Routing" not "Azure Routing" for the account. Azure routing is, IIRC, 12c per Gig. Internet routing is closer to 8c... Cant see an option for Servers, but I think Internet Routing might work for them too...
3: Any proof of this? I do know there is a limit of 256Mbyte for free account files, but not sure about what the limits on Pro or Business accounts...
I love cloudflare, but related to your last point can they still be called a startup? They’re a publicly traded company whose stock price is currently in freefall.
CloudFlare's stock price changes are related to narratives from the past 2 years rapidly fading, like the retail investing world finally learning what a CDN did right at the start of COVID, and expectations of the interest rate environment shifting. AFAIK there has been no material news to justify what's happening to its stock recently, other than many large players rushing to exit extremely crowded trades they all had in common.
There is currently a widespread fear that quantitative tightening could arrive very rapidly, with disastrous effects on high multiplier tech stocks and other debt-fuelled businesses. They've all been getting hammered in uniform since at least December.
For those following along, the aws cloud front price for 70 TB / 100m requests runs $5,400 or so. Discounts available if you will do a one year term - say 20%. Ie, they are in ballpark.
I can believe cloudflare is cheaper for sure, but it’s not your an idiot for using aws level I don’t think. And aws is printing money.
I once paid $3 per GB (6 figures per year) way way back so this all looks cheap to me.
As the good old Tanenbaum quote goes: "Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway".
I work a lot with video production and mailing SD cards or driving around hard drives is still the best way to transfer data quickly and cheaply. The cloud is too expensive (both storage and egress) and P2P (Syncthing) is too slow with more than 2 contributors. Despite most of my regular team now having FTTH (except me!!), we still manually transfer media files and use P2P for the rest of the project.
Azure's pricing for data can be weird (or used to be.
Didn't they used to charge for transfers between resources in different AZ's in same region? And VNET peering I think was chargeable, so you'd have situations where within a region you'd be paying outbound AND INBOUND charges.
Anyways, the little I looked, the "free" bandwidth alliance data transfer really wasn't free from what I could see but I didn't look too hard.
Its not 12c per gig, its 8.75c per gig for Microsoft Routing... (at least from Europe and NA) and 8c for Internet Routing... Not a massive saving... Mind you, if you are based in South America, you go from 18.1c per gig down to 12...
For Us, mostly based in Europe, it would be a difference of around 995 per month on Azure Bandwidth costs, if all our bandwidth just came from Azure. Same for South America is 6 grand in savings...
They don’t list because they are still in the “what the market will bear” model of pricing.
My experience was their sales team talked a good game up until the time to bid, and then Akamai closed enough of the pricing gap to not make it worth switching. Cloudflare seemed like great technology but their pricing suggested they had already won the business. I can only imagine how they treat existing customers once there are switching costs.
To summarize your point, yes, Cloudflare is not entirely free for all. Cloudflare specifically states why they have this sort of model, from their S-1 [0]:
> Our free customers create scale, serve as efficient brand marketing, and help us attract developers, customers, and potential employees. These free customers expose us to diverse traffic, threats, and problems, often allowing us to see potential security, performance, and reliability issues at the earliest stage. This knowledge allows us to improve our products and deliver more effective solutions to our paid customers. In addition, the added scale and diversity of this traffic makes us valuable to a diverse set of global ISPs, improving the breadth and economic terms of our interconnections, bandwidth costs, and co-location expenses. Finally, the enthusiastic engagement of our free customer base represents a “virtual quality assurance” function that allows us to maintain a high rate of product innovation, while ensuring products are extensively tested in real world environments before they are deployed to enterprise customers.
That's the value they get for providing the service for 'free' for quite a lot of websites that aren't moving real data volumes. For millions of websites, this is a sweet deal for them since it still saves them hundreds of dollars in bandwidth a month while they provide enough value to CF indirectly to not be considered a loss-leading customer (in which the customer would get "the call" saying they need to start directly paying for their usage of CF).
So is it an opaque, customer-hostile pricing strategy? Sure, but even image boards doing petabytes a month serving images are within the realm of never needing to pay for their direct usage of the service since CF still sees that as a worthwhile marketing and intelligence source. It's only when you actively cause them to lose money (such as streaming video in eg. the Oceania region) or start proxying raw binaries on a domain that proxies nothing else [1] that CF doesn't see a reason to keep your domain around as a charity case.
Haha. 21x for Telestra. Total crazy. They must be getting it from someone though, but that is abysmal. Good on cloudflare for detailing it - it's not just cloudflare who is impacted, the fees charged by others are also very high in these areas.
Point 3 is very true, however as far as competing on price, CloudFlare was a fraction of the cost of cloudfront when we had priced it out, despite the nontransparent pricing.
This includes the CDN product, workers, security services like Access and Teams. Only the Enterprise plan and other products require negotiation, but the majority of customers are easily covered by the standard options.
> "high bandwidth (games / software updates / video streaming) that is going to violate some type of clause"
The clauses are publicly stated in their terms. The CDN is for typical website assets and not large files, but many sites are running TBs of bandwidth for free or with the regulars plans though. Large file delivery requires different technical setup from small/dynamic requests, and there are CDNs specialized for that. You can also rent unmetered servers or interconnects yourself and get much better rates instead of using a CDN.
But I used to play with the "free" and "unlimited" players - and it's almost always the same.
The network quality is crap.
The network cannot sustain bandwidth during periods you need it.
There are hidden TOS stuff (sometimes made up) that they grab you with if you start using "too much" of the "unlimited" free service.
They are losing money so cannot sustain the pricing.
Pricing is opaque or not aligned with underlying costs.
Or they just flat out will not disclose their "80x cheaper" pricing because it's a total lie that it's 80x cheaper.
In short, I'm tired of that as I've gotten older.
Let's take cloudflare. If I egress 1PB of 8K video 100% into India per day, my guess is their per minute "upfront" stream pricing generates a call from some sales dude. Oh, you violated xx/yy/zz, we can't support that etc. Ugh. Same with their CDN stuff. If I start doing 100% egress into South Africa - wham - some weird call. I don't even have to look anymore to know how this goes.
If I strung together all their bandwidth alliance folks and started moving AWS Snowmobile data around per week (100PB per week, distributing it using the bandwidth alliance backbone for "free", without using other significant paid services) you get the call.
I ask simply, where is their 80x cheaper CDN pricing? OH, asking for pricing, you must work for AWS. It's pathetic really - that's the response? Point me to the webpage for this.
Before any discounts for committed use (normally 12 months or more) AWS Cloudfront goes down to $0.02/GB. The supposed 80x cheaper Cloudflare price is then in the $0.0002/GB range. IF you actually believed they had their costs low enough to offer this, they would clean up. Instead it is almost certainly a lie - but I'm happy to be proven wrong. If you do have needs this cheap, you might look at building your own CDN (see Netflix and friends).
The real irony is they do blog posts using AWS pricing, but don't release their own. I use cloudflare, I think they are great as a hobby level user of their stuff. But stuff for work goes to GCP (in past) and more recently AWS and Azure (which has a big in these days with office 365 / directory / remote desktop stuff into bigger players despite what may be underlying tech weaknesses).
The amount of sales calls and PowerPoints are crazy with CloudFlare. AWS makes it soooooo much easier to spend money and their sales are so less naggy.
> The Services are offered primarily as a platform to cache and serve web pages and websites. Unless explicitly included as part of a Paid Service purchased by you, you agree to use the Services solely for the purpose of (i) serving web pages as viewed through a web browser or other functionally equivalent applications, including rendering Hypertext Markup Language (HTML) or other functional equivalents, and (ii) serving web APIs subject to the restrictions set forth in this Section 2.8. Use of the Services for serving video or a disproportionate percentage of pictures, audio files, or other non-HTML content is prohibited, unless purchased separately as part of a Paid Service or expressly allowed under our Supplemental Terms for a specific Service. If we determine you have breached this Section 2.8, we may immediately suspend or restrict your use of the Services, or limit End User access to certain of your resources through the Services.
If this was truly acceptable and not in some grey area, why doesn't Backblaze simply route all downloads through Cloudflare by default, rather than having each individual customer go through the hassle of setting this up?
The page says that the Bandwidth Alliance means that partners will charge less or no egress to cloudflare.
I'm not seeing it saying anything about different rules applying regarding CloudFlare ToS such as "2.8 Limitation on Serving Non-HTML Content" to Bandwidth Alliance sources.
But has that been said somewhere I'm not seeing? Would love to see it!
If you only use this for "standard" CDN assets (like pictures that are part of your website styling rather than as an image host) and you also host your website on Cloudflare, I think it should be ok.
> If this was truly acceptable and not in some grey area, why doesn't Backblaze simply route all downloads through Cloudflare by default, rather than having each individual customer go through the hassle of setting this up?
Because Backblaze makes more money charging you for repeated downloads?
CloudFlare are about to launch their own object/file store with native CDN called R2. They have even come up with a wonderful trick for migrating to it from S3 and it’s looks like it’s going to have an incredible generous level of free egress. I suspect that it will beat out the B2/CloudFlare combination once it launches.
Pretty common strategy. Facebook owns facebok.com, Apple owns aple.com, Google owns gogle.com, etc. Scammers try to buy domains that sound or type similar to legitimate companies (practice known as "typosquatting") and use this to fool customers into thinking they're on the real website. One easy way to combat this is to just pay up and buy all the domains that are "closely typed" to your real one and have them all redirect to your property.
About 10 years ago a friend showed me Google ad words and Google analytics. He could see miss-spelling searches sometimes ending up at his site. He purchased about 20 domains and adwords of different miss spellings and his traffic went up about 35% on his site.
What is your opinion on what's driving the movement?
I went in fairly big (for me) and sold near the top with a stop loss #gainz. Then bought again when it went down like 10% trying to 'buy the dip.' But it keeps falling!
I don't get it. I love their products and ambition. That was a huge bump and decline.
Not the parent but I think Cloudflare's free tier is too good and the paid offerings don't compete well if you're all in on a cloud provider already. It also may be the classic Cloudflare has "features" but doesn't have a "product".
Anecdotally, I have yet to work with a company paying Cloudflare substantial amounts of money.
I feel they are fairly early on in there “full stack” strategy. I believe they are moving quickly from launching MVPs for various products to expanding them as rapidly as they can. They are also purposely not trying to compete 1:1 with the incumbents with the “full stack”
“Workers” is brilliant, but the strategy with WASM on it is where I really think it is going to shine. The more languages and more of your stack that can be compelled to WASM the more you will be able to run on their edge. I truly believe WASM is the future in so many places other than in the browser.
I also feel the one part of their stack that’s missing is a transactional acid database (I hope for sql, but nosql would be fine). But it highlights a fault with the “edge first” strategy. Workers at the edge making multiple round trip requests to a central database is slower than running your app closer to the db. My wish is that they are developing some sort of edge replaced db that automatically launches read replicas at the same edge locations as where most of your worker load is. You probably have to have the main/master in a central location though. (I think Fly.io are doing something a little like this with Postgres)
Interesting. I like that whole view of a product. Like how do you describe Cloudflare's produce in one sentence.
Maybe something about 'large edge network cloud to reach your customer quicker' or something I'm not sure how to describe it in a compelling way. Maybe sprinkle on some web3 or distributed bs.
I think AWS, now Google is getting there, have that a lot of stickyness because it has all the services imaginable interconnected (not well sometimes but still). It's a platform not individual products. It's hard to pickup and leave, especially if Cloudflare doesn't have 1:1 replacements.
If someone offered you a burger for $500 you probably would not buy it, even if it was very good. Even if they suddenly discounted it to $350. It's just too much money for a burger.
This thing makes like $500m of annual revenue, not profit, whatever path to profitability one can imagine, hard to see how it can be "worth" $32000m. It's like 5000 years of earnings are priced in along with the assumption competitors don't try too hard for the same time period.
Basically every reality-based investor has long left the room. It's a meme stock. It's Dogecoin for devops kids. Underlying biz does not matter for meme stocks, one must look at what the meme lovers do, what trend they follow, etc. At the moment they're being moody, that's why it's down for now along with similar memetech stocks.
As a bonus they provide the captchas for all the crypto sites. When that sugar daddy leaves the room, LOL.
It’s gross margin and revenue growth are comparable to FB’s of 8ish years ago (gross margin sits around 80%. FB’s today is 90%, insanity). Even after it’s recent drop in valuation, NET is trading at a higher revenue multiple today than FB was at that time.
FB was probably undervalued 8ish years ago. But FB also had a clearer narrative around TAM and LTV than NET does.
The current price bakes in some biggish assumptions (1) NET’s TAM is not even close to saturated (2) LTV per customer will keep growing as NET adds on new services that existing customers will want to pay for. There are many valid reasons to believe this, if you follow their earnings calls.
That doesn’t mean today’s price is bona fide bad or good. Rather, it reflects Wall St’s faith in Cloudflare’s ability to materialize grand-but-achievable ambitions.
The earlier peak price had far too much growth priced in, it was not sustainable imo. When price outpaces value, it’s normal to see a correction.
There may be fundamentals at play related to their product offering/pricing too but pure growth companies with negative earnings have been getting routed with the announcement of Fed tapering and interest rate increases. That's the main thing driving Cloudflare's price down.
funny enough, B2+Cloudflare will still be cheaper, since Cloudflare R2 from the blog you linked is at the cost of $0.015/GB/month for stored storage, where as B2 is $0.005/GB/month for stored storage, and there is no download cost when using the B2+Cloudflare combo
Driven by the fact that using comparative adjectives in this context seems to be at best confusing, I thought I'd play around with some numbers to see if I could understand what the author was trying to convey.
First, it's important to clarify why comparative adjectives don't work here. Look at the following sentence.
- A $5 cookie is worth twice as much as a $2.50 cookie
Here, we establish the baseline (the $2.50 cookie), and express how much better the $5 cookie is.
We could also say the inverse:
- A $2.50 cookie is worth half as much as a $5 cookie.
What we can't do is the following:
- A $5 cookie is double more expensive than a $2.50 cookie.
We can make a statement (the cookie is more expensive), or share the factor by establishing the baseline (twice as expensive as), but not both at the same time.
The author tries to say 440% cheaper, (not mentioning storage or transfer) and I believe their logic was something along the lines of the below:
- 440% = 440/100
- Cheaper = Reciprocal of More expensive
- 440% cheaper = 1 / (440/100) = 100/440
- S3 is very difficult to estimate a 'per GB price' because it depends on where the data is being transferred to, various pricing tiers (S3, S3 intelligent, 'Glacier', 'infrequent access', 'One Zone', first 100GB free, egress through Cloudfront, and HTTP verb specifics). Fortunately, they have an estimate calculator, which gave me 9 USD for 100GB transferred out to the internet, or 0.09USD/GB for transfer.
- Storage is a different proposition, with Amazon telling me it'll cost 2.30 USD for 100GB as long as I never look at it (ie: 0.023USD/GB)
- Backblaze is much easier to reckon with, with $0.005/GB for storage, and $0.01
for transfer.
- Therefore, for transfer Amazon's 0.09 * (100/440) = 0.02, which is MORE than Backblaze would charge me.
- For storage, Amazon's 0.023 * (100/440) is indeed 0.005.
So, Backblaze does indeed seem to be [440%, reciprocal of more expensive] as Amazon for storage cost alone.
Then again, the title of the post says "practically free" and goes on to say it's actually $5 a month, so I took the rest of the figures with a grain of salt...
B2 itself is not $5 a month. That $5 service is a separate consumer-focused "unlimited" backup service from the same company. B2 just charges for storage and transfer. Yeah, the author was rambling a bit.
B2 costs me roughly 52 cents per month to backup 250GB of data. I'd say that if you have enough income to have an internet connection and generate enough data to want a cloud backup tool, it is effectively free.
Do you think "twice as expensive" is meaningless? Because by that same logic, if I start with "once as expensive" that's not a number and multiplying by two gives me a meaningless concept.
Once as expensive is a number, it's 1 x the cost. Twice as expensive is 2x the cost. Once as cheap might arguably be 1x the cost, but twice as cheap is undefined.
It seems like we're taking at cross purposes. I'm not trying to convince you that anyone says "once as X". I'm just saying "twice as cheap" is nonsensical. Because it's a multiple of something undefined.
My sarcasm detection is not great. Looks like a troll post. Maybe it is real? If it is real, 50% down would mean it lost half its value. 100% down it lost all its value. Even if it was up one billion gillion percent, if it lost 100% of it's value it would be worth zero. But, yes, Enron and Madoff and many brokers are not out to protect your investments
FWIW a partner of mine has been using this setup less for the CDN aspect and more for the free B2 egress, and a TB or two of large encrypted files (rclone crypt remote) have not resulted in any complaints.
Video streaming consumes dramatically more bandwidth than the HTML/CSS/JS/JPEG/etc files which Cloudflare's CDN is intended to serve. Bandwidth is cheap, especially for Cloudflare, but it's not free. Cloudflare's base pricing, and especially their allowance for free accounts, does not account for such large bandwidth consumption.
Cloudflare does have a product specifically designed to stream video. It is priced accordingly.
I've been wanting the browsers to support SRV dns records for quite some time now. Unfortunately, both firefox and chrome have bugs in the WONTFIX status. The reasons for these statuses given are poor imo, but the implementation could add an extremely interesting dynamic to the internet. It would be easy to set up vps around the world and let SRV records set up in a round robin manner control it.
Will it be as good as a bgp controlled one? Unlikely, but it would be 100% in your control and easy to set up, unlike going cloudflare.
This one might have a larger chance of being accepted, since it provides some things which HTTP standard bodies and HTTP client vendors want. It doesn’t (at this stage, anyway) provide for the load-balancing “weight” field from the SRV record, but it does support MX-style priority numbers, and also port numbers. Very interesting, to say the least.
Assuming this is for websites, what's the recommended way of handling cache busting with this kind of setup? The article has a slightly awkward paragraph that tells you to set long TTLs for everything and concludes with "tweak these settings", which kind of sounds like "draw the rest of the owl". Might be a good reason to use something that integrates all this (eg. Cloudflare Pages, since you're already using Cloudflare).
The usual approach is to add a value to the query string, like a last-modified timestamp or deployment version. Doesn’t require any setup on the CDN side.
I know every use case is different, so this may not matter for most. But in my experience, the performance of B2 was quite poor in comparison to other S3 compatible options. It seemed like the crux of the issue was slowness in their API. Generating pre-signed URLs for file upload or download was quite a bit slower, resulting in a poor user experience in my application.
If you're just using it to upload public static content for your site, this shouldn't be an issue. But if you're dealing with private content it probably isn't the optimal solution.
> Generating pre-signed URLs for file upload or download was quite a bit slower,
I'd be shocked if generating pre-signed URLs required talking to the Backblaze server, that should be something done totally in client code.
However, even with AWS S3, I discovered that the official AWS ruby SDK client was much slower that it could/should be at generating pre-signed URLs (and even just public URLs!), and was able to write my own ruby code to be significantly faster.
I'm curious for the specifics on what code you were using to generate pre-signed URLs for file upload or download with backblaze, and what you are comparing it to when you say it's a lot slower. It could be un-optimized code, but whose "fault" it is depends on what code it is, open source etc.
I would love to hear more. I had a gig where we considered S3 and B2 for serving very private files (financial records). With B2 I could not find a good way to ensure that, assuming the URL is known, a file could only be accessed by an authorized party.
I am in the same boat currently in my research. In fact I posted this gist link we are discussing. My use case is confidential data and this solution won’t work if I can’t guarantee access is limited to specific users around the country.
Is free, secure, and fast to much to ask? Joking of course.
I would love to hear some of the implications of this approach. I can already think that the attacker would know the size of the file, so that might give him information regarding e.g. the file type. It might be of interest to an attacker to know that most clients get files of a specific size, but a certain client gets a file that is much much larger.
Great idea, I just need to move my thought of encryption on the sending client to encryption on the destination server since this is infosec related and sending client could be compromised. Thanks for the help!
That URL is susceptible to being intercepted in transit, though, such as in an email. And validating the request via a redirect on my own server invalidates 2/3 of the reasons for using an external file host.
I've been using Cloudflare free in front of App Engine's free tier for a few years and it really seems too good to be true. I can go viral and scale to infinity, for exactly $0.
The other day one of the commenters under a post wanted to spend $20 per month to host a static website to handle HN traffic, I just LOL’d. I tried to be good and explain why it was so unnecessary but nah they would rather bring a canon to a water gun fight.
Or a fully static host, like Netlify, Firebase Hosting, CloudFlare Pages, etc. which have extremely generous free tiers. I got twice to top ~15 of the HN front page, including once briefly at #1, within a month ( the billing period) and got to only ~90% of my bandwidth allowance of the Firebase Hosting free tier ( which isn't the most generous one).
Not only is it cheaper, it's also with less hassle ( there's nothing to manage, you just send static HTML / code to build to static HTML).
Sure. Or a full cloud instance running on a free tier somewhere. There are tons of ways to host low traffic sites for free that can be made to scale to high traffic for free by slapping Cloudflare in front. As long as your site is mostly static.
In my case, I'm serving data that updates every 10 minutes but is otherwise identical for every user. Cloudflare handles that use case perfectly.
Yep pretty amazing what CF makes possible combined with free tiers of other services. That hadn’t quite clicked in my head until I read your comment. Thank you.
+1 for Bunny, if you cannot take advantage of a free tier somewhere and have to pay. Setup a client of mine on it for video and images, real-time image modifying/branding, smooth video playback (without using a special product) all using the equivalent of presigned URLs - for a surprisingly low amount of money each month. Lots of good stuff going on there.
They also have their own storage options that are tightly integrated to their CDN, but basic FTP and simple API for it still. No S3 or bulk options there yet. Promised S3 API support “soon” though
"if everyone does it, they won't let you do it" -the infinite cache life on CF only works because 99.99% of use of the cache stays in rational cache lifetimes. If everything had long persist, CF would begin cache ejection, and your cost goes up, or they change the ToS to forbid it.
Backblaze also has an S3-compatible API, so you should be able to use the S3 adapter with backblaze too, instaed of the third-party backblaze-specific adpater using backblaze-specific API. You can use either S3 API or Backblaze API with the same bucket, simultaneously even.
Haven't a lot of people on HN warned against using Backblaze for backup because of issues restoring from backup? I recall reading someone having to pay to have a physical drive sent to them to restore a backup because of speed/cost issues trying to restore over network.
You're probably thinking about their computer backup service. This post talks about B2, an object storage service like S3.
While B2 isn't as fast as S3 and some tasks are way slower (eg: deleting files), I'm able to max out my server's 1Gbps connection when sending or downloading files from their European DC (using rclone and a server also in Europe).
B2-Cloudflare was a great companion when Cloudflare's Bandwidth Alliance was a thing and outgress from B2 to Cloudflare didn't cost anything, which means that outgress was completely free since Cloudflare don't charge on outgress.
I have been very happy with DigitalOcean spaces for $5/month including 250GB storage, 1TB outbound, and custom domains. Seems like something more closely comparable to this set up so I am surprised it isn't part of the comparison.
I've done this by using Cloudflare Workers. I can confirm that I did not see any bandwidth charges at B2, although I've only tested a few GB of transit. Likely if you tried to host brew here, you'd get yelled at.
Is anyone building CDN service on top of IPFS? IPFS is certainly too slow for many applications but perhaps there are tiny niches where it makes sense today? More like tape drive storage?
In my experience their gateway is not great and serves captchas occasionally if you're requesting rare content they don't have cached. I ended up just running my own node at a provider supporting bandwidth alliance and putting cloud flare in front of that.
Quite some caveats. Minimum storage used per month is 1TB. That means, if you store 250MB of files on there you will still be charged for the whole TB. Also, you will be charged for storing those files for at least 90 days even if you only keep them for 1 day. Lastly, and this is a very important one, you only have free egress equaling the total amount of data you have stored. So, say you have 2 TB stored, you can only egress 2 TB that month.
If you start to exceed that limit they are free to suspend your account.
So it depends very much on your use case, but there were too many risks for my use cases.
Yeah, the bandwidth limit was a killer issue for my use case as well. I've heard from older businesses there were exemptions from limit for older enterprise customers but based on a couple sales calls I had with them, they no longer provide any exemption for bandwidth and there is no option to just pay for more bandwidth for any plan. The cost of accounting and having to always predict and significantly overshoot the amount of bandwidth needed easily outweighed any cost benefit wasabi gave for my use case. This seems like a strange limitation that they cant just charge me for max(storage_gb, transfer_gb) but it seems their infrastructure was designed with that limit in mind.
Because they give a lot away and post blog posts attacking AWS folks think they are cheap. But a lot of their free stuff may be in the marketing budget. The issue here is:
1) Their paid offerings for their core product, CDN, does not have a public pricing guide. That's a BAD sign if they claim to be competing on price. So we can pretty safely assume their paid service bandwidth charges are NOT the 80X cheaper than amazon despite their claims here.
2) They hype their bandwidth alliance. I tried following this up on Azure, it's basically all call a sales person / maybe you'll get a deal type stuff. Smoke and mirrors again.
3) If you actually try to use their CDN to serve something high bandwidth (games / software updates / video streaming) that is going to violate some type of clause which says that the free CDN is not REALLY unlimited.
4) Financially they spend a mind boggling amount on sales costs. Between sales costs and cost of sales I think they are negative relative to revenue.
BTW - If you try any of the other "free" and "unlimited" providers (for $5/month etc) you will also find out once you try to move some real data volumes (video etc) - it's all a lie over and over.
Same things with geo issues. Cloudflare claims to have "one" price globally. These are almost alwyas false claims. It's easy to test - setup a stream into a high cost area, stream a bunch of video there - you'll usually get a call over some tricky TOS issue they pull out.
In short, if you want pay and forget, there is a reason the big cloud players have a business still.