Clouds Are No Substitute For Competence
This morning I came across an AP article on Yahoo news calling out Amazon’s ‘cloud computing’ initiatives. The all-too-clever title “Amazon’s Hot New Item: its data center” caught my attention and I wanted to see the folks at AP take a crack at the topic.
The article seemed innocent enough until I ran into this quote from the CEO of Dallas-based startup Mile Meter:
During the first dot-com boom, he said, “It was a badge of strength to have as much as possible in house. “Now, unless that is your core business…it’s a liability.”
Hmmm… I think the dot-com era statement was a bit over the top. Realistically, there was no real alternative to in-house infrastructure management. The hosting business was itself developing almost simultaneously and it was a dumb idea to simply hand over management of your entire setup to a 3rd party. Of course, you also needed tons of expensive gear to run stuff too. Well, it turns out today you still need tons of gear (assuming you’re successful anyway) and while it’s cheaper, the gear itself is only part of the problem.
But it’s the second part of the statement that I have deep issues with (and I am perfectly open to the possibility of it being taken out of context). The idea that you have to be in the business of managing technology assets in order to justify not going the EC2/S3 route (and it must be stated that EC2, while tres-cool is still in beta) seems short sighted. In fact, it counters completely the idea that operational excellence delivered through a talented ops team and a good infrastructure is a competitive advantage. It certainly is for the big, established players in the business — including Amazon.com. Tim O’Reilly wrote about this a while back when discussing how operations is the new “Secret Sauce” of companies doing business over the web.
Of course, perhaps part of the value of the S3/EC2 offering is that you tap into Amazon’s ops excellence. From what I have seen from playing with it, all you get is virtualized storage and virtual machines which have some lifecycle tooling around them. In effect, hosted gear. No monitoring, no application management, no person to call to debug why your memcache setup isn’t performing, and perhaps most important… no SLA (not for EC2 yet, anyway).
EC2/S3 save you hardware and storage costs. Everything else regarding operations skill, capacity planning, scalability knowledge, and good setup techniques still hold true for any business which derives its revenue from the web.
If you like this post then please consider subscribing to our RSS feed. You can also subscribe by email and have new articles sent directly to your inbox.
Twitter 
February 3rd, 2008 at 12:23 pm
I couldn’t agree more, although Amazon EC2 is a massive jump forward in the ability to have power on demand, it is also a jump backwards for people wanting to use it for hosting web applications at the same time.
It solves some problems (e.g capacity management to a degree), but creates other problems:
No permanent ‘live’ storage (S3 is near live, but not good enough if you want to run databases on it without workarounds).
No instance continuity (if a physical machine dies, not only does your instance automatically die, and you lose the data that was on it at the time unless you’ve backed it up, but it won’t restart either).
No static ip addreses. (if your instance does die, either through you shutting it down or through a failure, there is a reasonable likelyhood you will lose the ip address you are using, and thus have to use a dynamic dns service to *try* and keep settings up to date).
No integrated services like load balancing, capacity management, etc etc etc.
Ok, I’m extremely biased (I work for FlexiScale the competitor to Amazon EC2), and we haven’t got all the things listed above working yet either, but we’re a lot further down the list than Amazon.
Heres to the next 18 months as the entire industry evolves!
Tony Lucas
May 15th, 2008 at 10:48 pm
Startups with products aimed at large scale adoption in a SaaS mode cannot afford to miss out leveraging cloud computing. All the points raised about no permanent storage, static IPs,SLAs etc hold good. Amazon is fixing these as we go along. As a startup do I wait till all these are done and dusted? I may not be able to afford to. Does it bring on more to my plate to manage? yes, certainly but with the right amount of automation and support persons in house it is definitely worth it. We are using this infrastructure in production mode from Jan 08 and yet to witness an instance loss or data loss. Touchwood.
May 15th, 2008 at 10:51 pm
Oops, forgot to include a link. If you need more details about how we leveraged cloud computing in our product please check out http://sunilabinash.vox.com
July 28th, 2008 at 11:12 am
Jav,
Very good points. As some of the other comments mention, there are alternatives to Amazon’s offerings at different price points. As this interview mentions (http://www.thewhir.com/features/072308_QA_John_Pozadzides_Layered_Tech.cfm), some of the success stories for cloud computing are not yet public to protect the competitive advantage realized by those companies who are reaping the benefits (Disclaimer: I realize John is a CMO and therefore his comments are subject to bias, however, companies like LT are doing very well.).
Jason (aka Mr Chargeback)
August 14th, 2008 at 9:48 pm
Well,
AWS is rather good thing.
Problem with IPs is partially solved by elastic IPs,
Problem with persistent storage their are planning to solve
Even while I don’t have support contract my questions
on AWS forums where answered rather fact and correctly.
As for FlexiScale,
Being in Europe is good. Having persistent storage good
too.
But support, can be better.
For example time to process issues like
‘instance xxx is not coming up, it’s not possible to restart it,
it not even possible to kill damned thing’ from may experience
took about 12 hours to process (it wasn’t holidays).
And current record of time it took FlexiScale to process request
is about a month (data sent, several enquires asking about state sent,
after about a month – answer that they will start processing and feels
sorry for delay)