My future in 'the cloud'

[reddog]

This ^^

If the firm are moving to cloud infrastructure, like we have here, then as said above the normal datacentre operations need to be performed. But, additional considerations have to be made like how you access the infrastructure\security, supplier (amazon) management and SLAs etc am0ongst others.

this is true, and i will start of by doing this. However my thinking is that in the end in order to have lots of automation, and hence lots of savings, companies won't simply replicate their existing IT infrastructure in the cloud (sort of IaaS), they will end up going down the PaaS route. Anyone got any experience of this, are most blue chips just at the stage of experimenting with lift bits of their infrastructure into the cloud without much modification?

[gables]

this is true, and i will start of by doing this. However my thinking is that in the end in order to have lots of automation, and hence lots of savings, companies won't simply replicate their existing IT infrastructure in the cloud (sort of IaaS), they will end up going down the PaaS route. Anyone got any experience of this, are most blue chips just at the stage of experimenting with lift bits of their infrastructure into the cloud without much modification?

I'd be interested in the state of play, or stages if you like. I imagine there'll be a spread of what's been done, it would be interesting to understand what that looks like.

We had a session with our provider who suggested that a lot\number of firms who had gone the outsourcing route were now re-employing technical staff but leaving the infrastructure in the cloud.

[fool]

It depends what you're doing but using a PaaS either removes versatility or won't do the same job as your IaaS. They tend to offer you hosted services for your data layers so you can outsource the problem, which is fine if you want to be a low paid monkey or a developer doing ops as a side gig happy choosing only a subset of blessed tech.

Anyhoo, the drive towards automation is about running apps at scale. It goes right back to my previous comment which was "cattle not pets" which means, yes you do need them sysadmin skills, but also you need to be able to write code to implement those skills on 1000s of servers that go away and come back at random.

If you want to take it a step further, start talking about service discovery and frameworks such as mesos or kubernetes. These will offer a PaaS like experience with the versatility of being able to roll and automate your own data stores.

[Jefferson]

this is true, and i will start of by doing this. However my thinking is that in the end in order to have lots of automation, and hence lots of savings, companies won't simply replicate their existing IT infrastructure in the cloud (sort of IaaS), they will end up going down the PaaS route. Anyone got any experience of this, are most blue chips just at the stage of experimenting with lift bits of their infrastructure into the cloud without much modification?

I think at the moment a lot of companies are moving to SaaS products such as Office 365 and experimenting with bits of their infrastructure in the cloud (IaaS).
However, IaaS generally turns out to be more expensive than on-premise and with caveats regarding availability. E.g Azure will only honour SLA's on IaaS VM's if they are deployed in at least redundant pairs.

I think the future will be based more around PaaS as this offers greater savings and features e.g autoscaling of web servers.
This will require a lot of redesign for existing applications before they can be moved to this model.

I would say getting up to speed on Azure / AWS combined with traditional skills would definitely put you in a strong position going forward.

[fool]

However, IaaS generally turns out to be more expensive than on-premise and with caveats regarding availability. E.g Azure will only honour SLA's on IaaS VM's if they are deployed in at least redundant pairs.

I suspect this is generally only true when doing a fairly shallow comparison.

Sure, it you've got a single rack maxed it's going to be cheaper than the compute from AWS. That's not all you're paying them for though.

Can you do 3 datacentres, massive interconnects, autoscaling and failover with spare capacity and all the things and more importantly people those things entail? No, you probably can't, not unless you're very big and at that point amazon will cut you a deal.

Even if you're being cheap and don't care that much about the reputation hits that downtime will give you, the dev/ops time saved in itself probably pays for AWS.


P.S. Don't take this the wrong way, it's not intended to disparage. I'm just attempting to point out where the value actually lies as otherwise it's hard to decipher as buzzword overusage muddys the water.