Email or username:

Password:

Forgot your password?
Jeff Geerling

"But you can't make local infra as resilient as the cloud!"

Pfft, I think we have a decade of proof that statement is false. Local's also faster, and cheaper, if you invest anywhere near the same resources and don't just have a revolving door of contractors.
mastodon.bsd.cafe/@stefano/113

12 comments
gary

@geerlingguy i gave up on bonded 2.5 nics - great learning exercise though

gary

@geerlingguy local is faster with compute and ai if you make investment in fast network - I think network is big weak link for most smb - smb will still need cloud, saas, vps options but they should try to bring more in house if possible #tech consulting #enablers #fbi approved

rampantKitteh

@geerlingguy BuT sCaLaBiLiTy!!11

as long as people would commit to subscription-paywalled based services( this is: recurrent payment for access to the same service) that long cloud will be promoted as THE way to go.

Dr. Quadragon ❌

@rampantKitteh
> BuT sCaLaBiLiTy!!11

Yeah, scalability what? Buy more hardware and scale to your heart's content.

@geerlingguy

rampantKitteh

@drq @geerlingguy don't ask me, ask these who praise cloud computing as being $CORPORATE_BULLSHIT_GOES_HERE

Шуро

> Buy more hardware and scale to your heart's content.

If only it was so easy...

Шуро

> as long as people would commit to subscription-paywalled based services

Do support contracts for enterprise level hardware count as such?

SAN equipment, storage systems, servers, etc... are all coming with these and quite often you can't even download drivers and firmware without active subscription. Replacing failed components is a hassle or impossible as well. Everything is licensed, sometimes with recurring costs.

Some of it can be somewhat mitigated with turning to open source solutions but usually it just offsets costs from vendor to your own staff as you need to have more and better qualified sysadmins to support them.

> as long as people would commit to subscription-paywalled based services

Do support contracts for enterprise level hardware count as such?

SAN equipment, storage systems, servers, etc... are all coming with these and quite often you can't even download drivers and firmware without active subscription. Replacing failed components is a hassle or impossible as well. Everything is licensed, sometimes with recurring costs.

Brian

@geerlingguy I understand the broader point you’re making, but using a post where the local infra was the component that failed is a bad example. At least it was a stateless resource (Internet connectivity) that could have easily been replicated.

DrYak

@geerlingguy cue in people complaining that cloud can better be made nuclear-proof(*).
And thus completely miss the point of the hospital story.

(*): for some very weird corner cases which haven't actually been really tested.

Linker3000

@geerlingguy Agreed. Biz I managed for 3 years was all on-prem with redundancy, a local spares stock and 2 or 4 hr call-out for critical outages not fixable from spares. Cost was reasonable. Our only major outage was an HP battery-backed cache controller that took itself offline due to a spurious cache confidence error. We had the board swapped and the server + storage array back online in 12 mins. My successor moved many systems into Cloud and suffered the uptime and cost consequences.

Nanook

@geerlingguy "The Cloud" is just someone else's computer. There is no reason your systems can't be as resilient, all a matter of how much effort and money you're willing to put into it.

Dr. Quadragon ❌

@geerlingguy

> I think we have a decade of proof that statement is false

In some cases, very literal proof, as in a decade of uptime on a local server.

Go Up