2
molaram
18d

I thought this cloud shit was supposed to be more affordable and easier to manage and so on so forth.

After all the drama in the last months we just finished moving all our shit to bare metal. The bill is half of what it used to be but now we have some 95%+ capacity in reserve and waaaay less time wasted on putting all services together. It feels just like getting in a 911 after 7 years spent in a fucking suzuki.

So fuck AWS, fuck Google cloud and even though I wouldn't touch them with a barge pole, fuck IBM and Azure too!

Comments
  • 0
  • 0
    It can be both more affordable and easier to manage, in many (I'd say even most) cases, so long as you pick the best tool for the job and you use the architecture to your advantage. In some cases you'll trade the expense for the ease of maintenance though.

    Sad to say though I've seen a more than a few examples of "hey let's just put everything on EC2, why doesn't it just automagically scale everything?" and "hey I'll just stick all my stuff in lambda because that's the new hotness, wait why are my cold start times so awful?!" etc. In those sorts of scenarios you just end up with the worst of both worlds - expensive, difficult to maintain and a useless layer between you and the bare metal.
  • 0
    Cloud offers e few real advantages.

    If your architecture is good its very easy to scale up or down as needed.

    They usually can offer better ddos protection.

    They most likely can offer much better redundancy and local servers in many countries.

    If you fo not need any of those things you can probably get away cheaper, if you also have the expertise to set it all up.

    Not every company has that, in which case that can be a big cost to.

    So the answer is never a simple yes or no.

    Setting up my own server in the basement is dirt cheap, but I trade away a lot of features offered in the cloud.
  • 0
    I don't know about every possible use case out there but in our use cases we've discovered the cloud presents a variety of solutions to problems that don't really exist in real life.

    Very few companies run operations the size of Facebook, Google, Amazon and so on.

    Take autoscaling for instance, it's basically a solution to the high cost of cloud computing resources, where one is forced to scale up or down based on demand in order to avoid paying such fees.

    So first they up the pricing & drop performance, then they force you to break your application into scalable components which come with a plethora of challenges and problems, and guess what - for most of these they offer service X and Y at extra cost, but you have to foot the bill for all the work required to stitch them all together.

    Combine that with the tendency to over-engineer the crap out of every simple fucking thing, and you get artificially inflated demand for all sorts of resources and assets.
  • 1
    @Voxera I always argue that DDoS protection is nearly entirely useless. This becomes especially apparent if you know someone with a privately developed botnet and a truckload of custom exploits.
  • 0
    @PublicByte my current host also offer free ddos protection for all their datacenter. I'm guessing they have some dumb hardware they programmed to null route classes on abnormal traffic spikes - it's that simple.

    We also have cloudflare in front of the whole thing.

    About 4 years ago we had to deal with a rather large DDoS attack, first the AWS usage went to the roof, then cloudflare shut us down. Everyone was shitting their pants. The bosses gave all clients a call explaining the situation and we all took a couple days off.

    In our new setup what should happen is there would be no extra costs and all unknown IP classes will get null-routed indefinitely, but existing customers won't see any slowdown or outage. At worst, if the routers burst into flames the whole platform goes down again but no data loss or any extra costs.
  • 0
    @molaram the problem with ddos is volume and location.

    Large operators can reroute through other locations, that can be hard to do by your self.

    Of cause, a really large focused ddos is likely to kill any service.

    The real benefit of cloud is ease of setup as you can get a ling way without knowing much of redundancy and network.

    The other is scaling.

    Both can be done on your own if you have the skill, but actually very few companies has that skill readily available :)

    And hiring someone for it will most likely cost more than the cloud iv they are any good ;)
Add Comment