Another annoying cloud pitch I just sat through another cloud consulting services pitch from a major IT vendor. Some of it was under NDA so can’t talk about specifics, but that’s not really relevant to the point that has prompted me to write this. I guess my big problem is that I have spent the last ten years as an analyst looking at different disciplines …
The missing part of the article is that clouds come and go and eventually turn into rain - which nobody likes.
This is why it is best to avoid these 'clouds'. I have struggled like the author to work out what these Clouds are - that is other than outsourced hardware and service management with 'root' access available to me ... Otherwise known as hosting.
Its just a new catchphrase for an old idea (albeit an evolving and more complicated idea).
Look up the Eli Lilly case study .. they were able to spin up a 64-machine cluster on Amazon to do some bioinformatics sequencing .. it took 20 minutes to do the run and cost them a total of $6.40.
To do this the traditional way would have taken weeks and tens of thousands of $.
This kind of thing was not possible before the "Cloud" no matter what Larry Ellison and all the other "this cloud stuff is nothing new" people say.
"To do this the traditional way would have taken weeks and tens of thousands of $."
Just out of curiosity, how many weeks, and how many thousands of dollars, did it take to get everything together to run the batch on amazon hosted servers?
Some of us are old enough to remember submitting jobs to the/a processor and waiting a measurable period of time for a result. Most of us don't want to return to those conditions.
Not quite true. IBM has at least one computer & OS that will run on all 60 some odd processors.
However the cloud tends not tend to help processes that have a need for say 64 (or almost any number for that matter) as in order to take advantage of the processors they can only run in parallel
if the output of one process needs to be input to the second etc etc. If a program can start say 64 processes that is fine and it will help but todays computers cannot do parallel work unless the work is really independent from the result from any other processes.
Example you program the computer to multiply 2 * 2 * 2 * 2 * 2 * 2 * 2 * 2 etc etc it can only do that on 1 computer at a time as the results are needed before the second calculation can start, etc etc.
@jimt222
What you mentioned is nothing new and this was the point of the article I believe.
The martketeers etc are simply rebranding existing technoligies and methodologies as something new and exciting - of which it is usually neither
Companies have been renting external computing capacity for decades. I recall doing this in the 1980s at ICL.
Most recently I beleive it was referred to as outsourcing.
Dr Funk; I think that there are critical differences between the outsourcing that has occurred for decades and the spontaneous availability of computing resources on the 'real' cloud.
The bio company that jimt222 referred to no doubt decided on their solution, put in some credit card details and had access to the systems to power their work within minutes. Critically, when they finished the processing the charging stopped.
With ICL; did the outsourcing of your compute requirements start with the courting by sales types, the to and fro of contracts, minimum spend (though with an element of pay for the level of resources consumed) and importantly minimum contract length - at least a year if not a lot longer? How long did it take between you deciding to outsource and having the systems available? More than an hour, more than six months?
The Cloud is the just the latest buzzword, liberally applied to any compute farm infrastructure; hence so much of the traditional outsourcing market being resold as if were something new.
There is something new here; a new level of flexibility.
"Companies have been renting external computing capacity for decades. I recall doing this in the 1980s at ICL."
Exactly.
In the '70s, I used Tymeshare ... IBM has been doing the same thing since the '50s.
The more things change, the faster marketing separates fools from their money by selling ancient concepts as "new" ideas ...
Clouds indeed have been around for quite some time (at least in the IBM world). As long as the system is set up correctly any job can run almost anywhere (and successfully) from the same data center to 50+ kilometers away. Granted not too much more than that but it is being done today in the US and Europe at least (not sure about other locations) but there is no real technical reason why it couldn't be anywhere.
Online systems can indeed fall back almost instantaneously from any system to another system. There isn't any magic in it it just takes planning.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
