422
A single point of failure triggered the Amazon outage affecting millions (arstechnica.com)
submitted 1 year ago by to c/technology@lemmy.world
75 comments fedilink hide all child comments
 

We need more cloud services.

you are viewing a single comment's thread
view the rest of the comments
[–] 2 points 1 year ago (6 children)

We don't have to. It is entirely possible to engineer applications and services in a way that they're not dependent on any one cloud service, while also using cloud services for IaaS. Netflix famously does this, and sure enough Netflix experience no service interruptions during this latest outage despite having a large AWS presence.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 6 child comments
    •
  • [–] 1 point 1 year ago (5 children)

    If we want a truly robust system, yeah, we kinda do. This sort of event is only one of the issues with allowing a single entity to control pretty much everything.

    There are plenty of potential issues from a corrupt rogue corporation hijacking everything to attacks to internal fuck-ups like we just experienced. Sure, they can design a better cloud, but at the end of the day, it's still their cloud. The Internet needs to be less centralized, not more (and I don't just mean that purely in terms of infrastructure, though that is included of course).

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 5 child comments
    •
  • [–] 1 point 1 year ago* (4 children)

    If we want a truly robust system, yeah, we kinda do. This sort of event is only one of the issues with allowing a single entity to control pretty much everything.

    What I'm advocating for is the opposite of "allowing one entity to control everything".

    https://en.wikipedia.org/wiki/Chaos_engineering#Chaos_Monkey

    Read about it dude. Netflix has a large presence in all major cloud providers (and they have their own data centers), but has a service whose uptime is NOT dependent on any one of those hosting environments. The proof is the pudding - Netflix service did not go down in the recent AWS outage, nor in the last one.

    All of that can be achieved WITHOUT completely abandoning public cloud services and having to completely host all of the hardware for their services.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 4 child comments
    •
  • [–] 1 point 1 year ago* (3 children)

    Yes, Netflix had their own infrastucture in addition to other multiple redundant cloud services for their CDNs: You're kind of proving (part of) my point?

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 3 child comments
    •
  • [–] 1 point 1 year ago (2 children)

    You're kind of proving (part of) my point?

    How? Their reliability would exist without that. There's nothing inherent to their own data center that makes their setup that much better. Having a distributed system across multiple cloud service providers means your actual chance of downtime (here I mean inverse of uptime) is their individual chances of uptime multiplied by each other. In other words, they all have to go down for your service to fail. The catch is you have to use only commodity IaaS and PaaS, nothing proprietary to one CSP.

    For smaller companies especially, in terms of pure reliability, there's no reason to think that they would be better at running a high availability data center than Microsoft or AWS or Google.

    Parallel distributed architectures give you the advantages of using public cloud (not having to physically manage your own data center) without the disadvantages (dependence on any one cloud vendor), while also potentially increasing your reliability beyond the reliability of any one of your cloud vendors . That is why Netflix is so rock solid.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 2 child comments
    •
  • [–] 1 point 1 year ago* (1 child)

    You really don't see the risk of having no data centers you actually control as an organization? Maybe I misunderstood your initial statement? At first it sounded like we kind of agreed with each other but didn't understand that was the case at first.

  • permalink
  • fedilink
  • source
  • parent
  • hideshow 1 child comment
    •
  • [–] 1 point 1 year ago*

    You really don't see the risk of having no data centers you actually control as an organization?

    This really depends on what you think you're getting from having your own DC. Is it reliability? Flexibility? Control? What are your objectives?

    There's some argument to be made to have some locally hosted stuff for some flexibility and control. And in some niche cases the pricing of public offerings doesn't make sense.

    But as I said, if you're building your own data center for increased reliability then 1) you're necessarily assuming the premise that you're going to be better at managing DCs than Google, Microsoft and AWS which I think in reality would be hard to prove let alone do, and 2) is hard to justify considering you can distribute workloads across multiple data centers already (as proven by the Netflix example) so that your reliability isn't limited by any one vendor.

  • permalink
  • fedilink
  • source
  • parent