OK, I'm having a hard time understanding this "cloud" stuff

B

Beastie7

As far as I can understand, the "cloud" is more of an abstraction of how to deploy services at scale, right? And being able deploy at scale requires consolidating services I'm assuming? With all this "Hadoop", "Open Stack", and "Docker" stuff, it seems like people are trying to a take bajillion servers, stuff them together, and then tailor that cluster to a particular workload.

So... why not just use a mainframe? Isn't that (scaling up, I suppose) similar concept? Is x86 commoditization a cause?

I came across this thread and it got me wondering. Why isn't "scale up" so popular? Also, it seems every few months or so I keep seeing a damn new something-as-a-service popping in the blogosphere. Like "Metal-as-a-service" (MAAS). The hell does that even mean? Lol
 
I don't think scale has anything to do with it actually. The "cloud" simply means it's a service (whatever that may be) that's available everywhere. It's really just a buzz-word that simply means it's always available from everywhere. If I set up a FreeBSD box with a couple of bhyve VMs on it and rent it to you I can call myself "cloud provider". Anything on the internet can pretty much be categorized as "in the cloud".

As for the lovely new "something-as-a-service", yeah. New ones keep popping up all the time. Again, a lot of buzz-words.

https://en.wikipedia.org/wiki/Cloud_computing#Service_models

Why do it? Well, it's a matter of costs really. I myself have a couple of VPS' I rent. That's cheaper that actually buying the hardware and move it to a hosting location. Not only that, there's maintenance too. Harddisks tend to die. Power supplies too. With a VPS I don't have to worry about it as that's being dealt with by the provider.

You can go a step above that and rent a completely furnished server with some software on it. The provider takes care of maintenance, patches, hardware etc. That means a company doesn't have to hire someone to take care of it. Smaller companies simply do not have enough work for a dedicated "Tech" guy (or girl). So renting it is more cost effective.

Each step of IAAS, PAAS, SAAS etc. just adds another layer of management that's taken care of. Companies like that, it means they can focus more on their core-business instead of IT.
 
A common understanding of cloud architectures is that they support "elasticity". That means that if you add a further bootstrapped node, all the architecture will expand and gain efficiency. It's essentially a distributed system simulating one undistributed system (you don't need to code every single agent uniformly, it will automatically set up everything for you).

Of course the whole thing is a dream and underneath all this cloud facade you'll probably discover that it essentially consists of many massive farts.
 
nakal said:
Of course the whole thing is a dream and underneath all this cloud facade you'll probably discover that it essentially consists of many massive farts.
Click to expand...
Farts as a service.
 
SirDice said:
I don't think scale has anything to do with it actually. The "cloud" simply means it's a service (whatever that may be) that's available everywhere. It's really just a buzz-word that simply means it's always available from everywhere. If I set up a FreeBSD box with a couple of bhyve VMs on it and rent it to you I can call myself "cloud provider". Anything on the internet can pretty much be categorized as "in the cloud".

As for the lovely new "something-as-a-service", yeah. New ones keep popping up all the time. Again, a lot of buzz-words.

https://en.wikipedia.org/wiki/Cloud_computing#Service_models

Why do it? Well, it's a matter of costs really. I myself have a couple of VPS' I rent. That's cheaper that actually buying the hardware and move it to a hosting location. Not only that, there's maintenance too. Harddisks tend to die. Power supplies too. With a VPS I don't have to worry about it as that's being dealt with by the provider.

You can go a step above that and rent a completely furnished server with some software on it. The provider takes care of maintenance, patches, hardware etc. That means a company doesn't have to hire someone to take care of it. Smaller companies simply do not have enough work for a dedicated "Tech" guy (or girl). So renting it is more cost effective.

Each step of IAAS, PAAS, SAAS etc. just adds another layer of management that's taken care of. Companies like that, it means they can focus more on their core-business instead of IT.
Click to expand...

Ah, I see. So it's more a cost-induced practice than a new shift in technology

nakal said:
A common understanding of cloud architectures is that they support "elasticity". That means that if you add a further bootstrapped node, all the architecture will expand and gain efficiency. It's essentially a distributed system simulating one undistributed system (you don't need to code every single agent uniformly, it will automatically set up everything for you).

Of course the whole thing is a dream and underneath all this cloud facade you'll probably discover that it essentially consists of many massive farts.
Click to expand...

Yeah, somewhere I read "scaling-out" horiztonally provides more redundancy and compute per dollar in comparison to "scaling-up", whereas scaling up provides more stability and security for stuff like databases. I guess each depends on the workload you're putting out.
 
Yes, there are various ways to scale your infrastructure and they all depend on what you're actually doing with it. For example, if you have a heavily loaded website it makes more sense to add a load-balancer and a bunch of "small" web servers instead of scaling to one massive web server.
 
Yes, this works mostly, because web is agent-based (distributed system), and even better when you obey its basic principles (REST). On the other hand the principles of web are usually neglected.
 
Beastie7 said:
As far as I can understand, the "cloud" is more of an abstraction of how to deploy services at scale, right? And being able deploy at scale requires consolidating services I'm assuming? With all this "Hadoop", "Open Stack", and "Docker"
Click to expand...
Hadoop and it's counterpart Sparc is just a virtualization layer on the top of Apache web/application server which has its own file system Hadoop does. Docker is just a poorly implemented sandbox. Think of vkernel from Dragon Fly. Linux containers are supposed to be Linux answer to Solaris zones and FreeBSD jail. Openstack is just another abstraction layer on the top of level one hypervisor for example Xen Dom0
 
Docker does not look like a virtualization and neither like a security middleware. It rather looks like someone is missing easy-to-use 3rd-party application distribution mechanism for Linux (only), similar to Apple's dmg images or Microsoft's install.exe.
 
You must log in or register to reply here.
Back
Top