I've been following cloud computing since it was simply called  
virtualization/distributed and there was a wonderful tool with linux,  
to manage and expose physical hardware as a whole, where you could  
deploy then VM's inside. OpenMosix, was his name, now dead. Saw Xen,  
KVM, VirtualBox, UML, qemu and OpenVz all flourishing. Hell, I even  
got into LPAR while working for the - one and only - portuguese public  
bank, in Linux on Mainframe.

Now, with the 2.0 advent and service-this, service-that, there's a lot  
of companies going into the "cloud"... To sysadmins this is just  
another way of doing stuff. We're giving up on physical hardware  
management and entering the virtual world, leaving the physical world  
to our hardware supplier. It seems naturally. Yet, these cloud  
computing servers still need sysadmin work - will talk about it in  
another post.

I'm personally fan of AWS - Amazon Web Services -, since it has all  
the tools and solutions that every service needs, when you thing on  
scalability, availability and pay-as-you-go. But there are more for  
storage, database, content delivery, etc, you name it. AWS is the only  
one that has the flexibility for people to upload their own machine  
that they build. For sysadmin perspective this is rather important,  
since there's some tools to help us manage "intra-cloud" (cloud  
computing in-house, local) or "inter-cloud" (internet/public cloud  
computing). You can read further details on my codebits talk: 
http://www.vitordomingos.com/talks/codebits_20081112.pdf

Still, Cloud Computing is a new great tool for sysadmins, but its the  
holly grail for developers. On a private cloud computing mailing list  
someone asked why do we need cloud to scale from 1 or 2 machines to  
10, 20 or 100, on demand, automatically or just with a few clicks.  
Well, the answer lies with the development team and with regular  
service scale. This last one is rather obvious, the service needs to  
scale horizontally, vertically or both, when we reach peeks. But, the  
first one is using cloud computing scalability to hide bad software  
development. Heck! If software doesn't scale by itself, let's just  
throw it some hardware to solve the problem - it's cheap, pay-as-we-go  
and on-demand. This is why cloud computing is a stick of two nozzles.  
It's being used to address development screwups rather than ease  
sysadmin job.

One of Guy Kawasaki's startup rules is "don't worry, be crappy". I'll  
just complete it: "don't worry, be crappy, there's cloud computing".

//your beloved BOFH
//VD
_______________________________________________
tce mailing list
[email protected]
http://lists.paradigma.pt/mailman/listinfo/tce

Responder a