Anybody migrating on premise infrastructure to the cloud?

HouTxPilot82

Pre-takeoff checklist
Joined
Nov 6, 2014
Messages
286
Display Name

Display name:
HouTxPilot82
Seems like we have quite a few IT folks on this forum with a wealth of experience. I'm curious if anyone is working for a company that is moving their applications/systems into the cloud - specifically using Infrastructure as a Service. I'd love to hear some feedback from people who have gone through this transition. I don't care if you've gone to Azure, AWS, Google or some other provider. I'm interested in hearing about approaches, challenges, the good, the bad, and the ugly - what has worked, what hasn't worked, what would you have done differently.

I know some are convinced this is a bad idea from a security standpoint and would never trust a public cloud infrastructure. Others are convinced that's not the case. I'd like to avoid this discussion getting bogged down in the merits of going to the cloud or not from that perspective. I really want to hear from people who have done it and can share some experience.
 
The most important part is to do your homework. Some people think 'the cloud'(tm) immediately saves you tons of money. If your apps are already designed in such a way that they are resilient and scalable, it can save you money, on the other hand if you have legacy apps that can't scale based on load and thus you always need the largest instance sizes and have to use all the same features as on hardware(highly redundant storage, etc) then it may not save you anything or might cost you even more.

I've seen it work the best with bursty applications which are amenable to scaling out and in based on load. In companies that have owned their own data centers with applications which have a fairly low base load they run in-house and then burst out when they have traffic spikes due to product launches or other events.

Another thing to consider is while you may save on staff in one area(the guys pushing your hardware around) it will likely cost you money on the other side to have developers who understand the way to best optimize things for the cloud.

Personally, I work for a consulting firm which does this sort of analysis for a living, but my day job is helping customers implement on-prem clouds to either move their applications from legacy hardware to their own hosted cloud, or to host a cloud for their customers. It sometimes saves them money by better utilizing their hardware and making their developers more efficient, sometimes it doesn't save them anything, some of them have requirements that they don't want to deal with getting their lawyers to sign off on a public cloud provider(HIPAA, PCI).
 
It depends on how your costs are now. The thing that always got me about this "centralization" or "cloudification" strategy enforced on us by corporate weenies is that it ended up costing us substantially more at degraded performance to do it. Of course, the corporate guys just drink the kool-aid proposed by whatever IT fad is coming along without any serious look at impact.
 
Thanks for the input, gentlemen. I agree with many of the points raised here. 'Bursty' applications that are able to scale horizontally are of course the ideal candidates for taking advantage of what IaaS offers - especially combined with some automation based on load. I agree many of the legacy line of business apps are not necessarily capable of this type of scaling. One thing I have noticed is the tendency to over-provision in the on-premise environment - partly because the 'vendor said so', and partly because of the 'just in case' scenario... the cost implications of this type of over-provisioning do become more transparent when you have IaaS.
 
I would say the most important thing to keep in mind is that "The Cloud" is just a marketing term for off-site servers. It doesn't matter if you use it strictly for storage or if you run all your applications off it, there are really only two things that matter...

1) What is the quality of their infrastructure? Their servers, their power supply and how they connect to the net. Service and security do matter, but it doesn't matter if they have Bill Gates and Superman on the job; if they don't have the quality of infrastructure to assure "Five 9" up-time. It also doesn't matter if they have multiple redundant MPLS-VPN connections to the 'net if THEIR provider is 2nd or 3rd tier. You want your "cloud" to be at a Tier 1 provider. Verizon for instance.

2) It also won't do you much good to go with a cloud provider that has a Terabyte optical link directly to UUnet if YOUR connection isn't capable of handling your own up/down traffic.


You cover those two bases and let a tech savvy lawyer make sure your service agreement meets your needs and you'll be fine.
 
Back
Top