Proof that software as service/cloud based, will never work for long term ...

[Bicurico]

OneDrive (and also other cloud storage services) comes with a non-obvious risk, i.e. a locked account. When MS's AI deems a file to be nasty content your MS account is locked and you lose access to everything associated with that account. There are already several reports of such cases (caused by false positives) and a quite cumbersome process to unlock the account.

I do have 100% of my content mapped to those 4 computers I use.
Having a account lock would not be fatal.

[madires]

Google did it again:

Google Cloud shows it can break things for lots of customers – not just one at a time: https://www.theregister.com/2024/05/20/google_cloud_network_outage/

[alm]

Google Cloud shows it can break things for lots of customers – not just one at a time

So what's your point? Servers you own don't go down? Or that servers you maintain yourself have a better up time?

[madires]

After reading the article and my post one page back you'll comprehend the context.

[CatalinaWOW]

I haven't seen one of my biggest gripes about cloud services mentioned elsewhere.  It affects me in IoT applications mostly - smart bulbs and switches.  I am in a rural area with relatively poor internet service.  Slow and many dropouts lasting from seconds to minutes.  It seems that these applications are intolerant of these interruptions and the devices go offline and stay that way even when internet service returns.

This behavior may make economic sense.  The relatively small customer base that wants this kind of device while living in a poorly connected area may be too small to warrant the investment in more robust connection strategies.  But it probably affects folks in somewhat better served areas.  The kinds which have throughput bottlenecks that slow response in evenings when every household is streaming multiple videos.  This behavior occurs in many suburban environments.

[alm]

After reading the article and my post one page back you'll comprehend the context.

So just laughing about Google making stupid mistakes, because they are the only ones that make stupid mistakes. Got it.

[madires]

Nope! If you go cloud don't rely on a single cloud service.

[soldar]

Google Cloud shows it can break things for lots of customers – not just one at a time

So what's your point? Servers you own don't go down? Or that servers you maintain yourself have a better up time?

Servers you own and control you own and control. Servers owned and controlled by others are owned and controlled by others who can deny you access for any reason or for no reason and you have no recourse. Servers you own and control would not need internet to access them so there is one less point that can fail. In general keeping control is better and more reliable than ceding control to others.

[PlainName]

Servers you own and control ...

While I am in agreement with the points you mention, there is a big downside that isn't mentioned: you own and control the servers. That means you have to create them and maintain them, and when they go wrong you have to fix them. It can be time-consuming (especially since they will need work just when you don't have the time) and aggravating (wtf is wrong with the bugger?!? It worked fine yesterday, nothing's change, no obvious issue...). And many potential users just aren't up to operating their own mini-datacentre for which they are solely responsible.

[alm]

Servers you own and control you own and control. Servers owned and controlled by others are owned and controlled by others who can deny you access for any reason or for no reason and you have no recourse. Servers you own and control would not need internet to access them so there is one less point that can fail. In general keeping control is better and more reliable than ceding control to others.

Unless you are big enough to have a fancy internet connection that allows static IPs, incoming port 25 access, sufficient bandwidth to deal with DDoS, etc, and build your own data center, you are likely to have to put your own servers in a third-party data center. Does it really matter if your data is lost because some Google employee blocked you (how many of the 500k or whatever the figure is GCP customers get blocked per year?), or because your server died due to hardware failure, software bug, a fire / flooding, etc? Do you control all of those factors? In all cases the data is gone unless you have an (off-site) backup.

Sure, given enough resources you can build more reliable infrastructure than the big cloud providers distributed across multiple geographic locations etc. They are not trying for ultimate reliability, just good enough. If you have those resources and need that kind of reliability, good for you. It sounds like a fun project. But I read more stories about "We lost data / were down for a long time because Jane ran the script against the wrong database" then "We were down for a week because there was a bug in GCP / GCP blocked us". So I'd say the risk of a platform like GCP are over-estimated here and the risk of the customer themselves messing up are under-estimated. It is much easier to mess up backups when you manage them yourselves than if you use a managed service.

[paulca]

There is still far too much "conflation" of many different types of beast which come under the term "cloud".

On one end of the spectrum you have nothing more than physical server co-locations in a DC.  At the other end of the spectrum you have consumer cloud services which are 100% coupled and you are 100% a slave to.  And everywhere in between.

[madires]

Spotify joins the brick-it club:

“Unacceptable”: Spotify bricking Car Thing devices in Dec. without refunds - https://arstechnica.com/gadgets/2024/05/pleas-for-open-sourcing-refunds-as-spotify-plans-to-brick-car-thing-devices/

[coppice]

Spotify joins the brick-it club:

“Unacceptable”: Spotify bricking Car Thing devices in Dec. without refunds - https://arstechnica.com/gadgets/2024/05/pleas-for-open-sourcing-refunds-as-spotify-plans-to-brick-car-thing-devices/

Well it was launched WAY back in 2021. Its ancient. You can't expect these people to support their products forever. I mean, who even keeps things when the year's warranty has expired?

[.RC.]

I mean, who even keeps things when the year's warranty has expired?

Well I do, but I am not wealthy.

[coppice]

I mean, who even keeps things when the year's warranty has expired?

Well I do, but I am not wealthy.

I was interested to see Apple's current ad campaign for the iPhone, emphasising how its built to be durable, gets regular software updates, and should last for years. Quite a change from the "oh, that's last year's model. That sucks compared with this new one" approach.

[Infraviolet]

Regarding the spotify Car Thing, there do seem to be some projects underway to liberate the hardware from its corporate cloud dependent software so that atleast something can be salvaged:
https://hackaday.com/2024/05/30/old-spotify-car-thing-hacks-gain-new-attention/

[paulca]

There is also a growing campaign trying to take game publishers to court for "deleting" games forever.

Basically publishers are turning off cloud services for old games rendering them unplayable forever.

[coppice]

There is also a growing campaign trying to take game publishers to court for "deleting" games forever.

Basically publishers are turning off cloud services for old games rendering them unplayable forever.

There is a more general case that in 2024 there is really no excuse for copyright material to become unavailable except in exceptional circumstances (e.g. it fell foul of the law). Huge numbers of books and recordings can't be obtained right now, regardless of how much you are prepared to pay.

[madires]

Snowflake (AI Data Cloud) and their customers (many well known brands) seem to face a worst-case scenario:
- Snowflake account hacks linked to Santander, Ticketmaster breaches (https://www.bleepingcomputer.com/news/security/snowflake-account-hacks-linked-to-santander-ticketmaster-breaches/)
- Kevin Beaumont:  I've now confirmed 6 major orgs running Snowflake cyber incidents (thread: https://cyberplace.social/@GossiTheDog/112536407633131499)
- Snowflake is in denial mode (https://community.snowflake.com/s/question/0D5VI00000Emyl00AB/detecting-and-preventing-unauthorized-user-access)

 

[DimitriP]

CDK Global provides retail technology and software as a service (SaaS) solutions that help dealers run their businesses and drive profitability. CDK currently serves more than 15,000 retail locations in North America.

Most of the time, except the last few days !!!

https://www.cbsnews.com/news/cdk-cyber-attack-outage-auto-dealerships-cbs-news-explains/

[NiHaoMike]

https://www.tomsguide.com/news/live/xbox-live-outage-july-2024

[Marco]

https://www.tomsguide.com/news/live/xbox-live-outage-july-2024

More a failed fallback than pure cloud dependence, otherwise turning off the network in settings wouldn't have helped. Likely sloppy programming which assumed that if internet is up then the services have to be up, a bit similar to Apple's online certificate revocation checks bricking Macs for a bit.

[mikeb1279]

There is still far too much "conflation" of many different types of beast which come under the term "cloud".

On one end of the spectrum you have nothing more than physical server co-locations in a DC.  At the other end of the spectrum you have consumer cloud services which are 100% coupled and you are 100% a slave to.  And everywhere in between.

This. We need to get better at communicating this to 'non-technical/non-computer' people too. Otherwise we are constantly trying to explain to people that while it may be great that you saved money after shifting your backups from local DC to AWS, becoming entirely dependent on a SaaS product for your core business is an entirely different ball game altogether.

I think SaaS people actually deliberately conflate the two issues to make sales easier tbh.....

[paulca]

Under the hood and behind the curtain it's honestly quite frightening how the cloud show is setup.

Leaving the architecture stuff asides, what "cloud" platforms basically are, is a collection of scripts and a database.

The recent terms to become buzzwords highlight this.

SaaS - software as a service
IaaS - Infrastructure as a service
IaC - Infrastructure as code.
"server-less services"
"code-less services"

It's constantly nesting itself inside more and more layers of what are just scripts basically.  Take kubernetes for example.  It is basically an optimistic goal orientated control loop.  Each iteration it looks at the situation, compares it to the script you gave it and makes the next move from A to B to eventually and continuously enforce that infra or set of pods is up.

However, writting the templates and yaml required to set that infra up was tedious, so...  they have scripts generate them for them using exotic and fragile template mecahnisms. (like Helm).

All of this then runs on virtualised hosts, VMs of different classes, all dynamically instanciated from "code".  So rather than writing 15 yaml files and configuring 15 configuration maps, instead you write a script to generate them all for you.  You tell it how many servers, what type, what networking, firewalling, software images etc. etc.  It generates all the yaml and yaml templates.

Look I love scripts, I used to love sysVinit and still dont like systemd, but it's not like that.  There there was a common language (sh) and a common well define singular platform.  Cloud is not like that, it's a sprawling array of different often competing tech.

It just looks like a big mountain of technical debt being driven ahead of the bulldozers of over-sell and under-deliver management.

[Kim Christensen]

Under the hood and behind the curtain it's honestly quite frightening how the cloud show is setup.

I initially read that as "clown show"...