Friday, June 11, 2010
Are there some weak points of cloud?
Do you like this story?
Well in any conference I walk people are concerned about security with cloud but I tried to prepare more with this list..Watch on!
- Performance. The recent problems with Twitter (“Fail Whale”) and Steve Jobs’ embarrassment at the network outage at the introduction of the new iPhone don’t exactly impart warm fuzzy feelings about the Internet and network performance in general. An SLA can’t guarantee performance; it can only punish bad performance.
- ROI: What has driven the initial stages of cloud computing (that is, software-as-a-service and, to a lesser degree, infrastructure-as-a-service) makes sense mostly as a function of short-term return-on-investment. And isn’t it funny how initial cost savings always seem to dry up over time?
- Market churn. Remember the dot-com boom? Something like that is taking shape in the cloud market, and will be followed by the inevitable slump (bust, crash, etc.). The cloud vendor you were counting on suddenly packed his tent and slipped quietly into the night? Now what? And who has your data? This does not go for big players like Microsoft, Amazon.
- Privacy. Don’t be lulled into thinking the cloud vendor assumes all liability for privacy problems.
- Security. Of course the biggest of all, this is still a problem. When email was introduced everybody raised the same concern but where are we now...so choice is yours..
This post was written by: Saumya Aggarwal
Saumya Aggarwal is a professional blogger, web designer and marketer. Follow him on Twitter
Subscribe to:
Post Comments (Atom)
Posts
1 Responses to “Are there some weak points of cloud?”
June 15, 2010 at 11:34 PM
I'm not sure for applicability of some of your points directly to the cloud computing.
For example, the performance problem is true for any limited resource (network, power etc.) that needs to be used. If you are using dedicated hosting or even your own CoLo/DC, your internet connectivity is dependent on some common resources outside of your control that are shared. So if you need networking for your application, it doesn't matter whether you are in cloud or outside - your resource is finite. (Now, hosting the servers next door to your client could address this - but that's rarely feasable)
ROI is exactly where cloud computing is a HUGE win for many types of load. In particular, the dynamic scale and 'limitless' resource allows you to use the cloud to cover for any spiked load. In a classical design, you have to either engineer for the peak-load (or high % of it) or give up business/responsiveness/availability at that time. The classical 'slashdot effect' would bring normal sites down/unavailable until you can find the appropriate resources or load decreases. After the spike, you are left with extra resource that is not useful anymore. With the cloud, you could scale up as required and scale back down once traffic departs. Similarly, bursty calculations (such as financial modeling, weather modeling, etc) may only be needed temporarily but required permanent investment without cloud.
The big problems that you don't mention are things like data transfer (in-out the cloud) which somewhat underlies your market churn problem and cloud 'location'. This is something that provides a problem if you are trying to address multiple markets and the cloud isn't close to where you want to be. Some of the cloud providers (Microsoft, Amazon) provide CDN functionality to bring the data closer, but you still hit the privacy issue you mention. But this is also a benefit for the cloud, as it is able to let you expand to locations that you don't have existing presence.
Another problem is code portability, which is growing as the clouds invent new programming models. For example, you wouldn't be able to host your Google AppEngine application using GQL anywhere outside of Google's DC due to the lack of external implementations/hosters. Similar applies to Amazon's SimpleDB and Microsoft's Windows Azure Tables. Due to the need to scale, the programming model in those cases is not compatible with existing technologies. However, Microsoft and Amazon do offer less-scalable (in size) options that are SQL based and able to migrate. With the right adapters you could migrate your Windows Azure or Amazon's AWS application with limited modifications as they support normal .Net/Java/C++ development. This is where generalized programming models (.Net and somewhat Java) help. I know this is an active area of research/development at Microsoft, with newer products improving the cloud/on-premise integration.
PS: I work at microsoft so am a bit more familiar with their offering.
Post a Comment