An interesting Anti-cloud argument from Anthony Lye of Oracle

Anthony Lye did a great briefing on BI and CRM that just wrapped up. 

One of the discussions was around key industries + applications that really couldn't utilize the cloud.

His argument went as follows:

• Retail POS (or mobile) needs to run transactions between the time of card swipe (or other ID) and the time of receipt printing (or other display) 
• Similar latency requirements exist in call routing or other personalization.
• Even if the customer can wait, if there is a call center involved, just adding the call center personell is expensive.
• Therefore, you need servers very close to the transaction (call center personell, POS app server, etc.)

I'm not sure I buy this.

One can better handle peak volumes and QOS by over provisioning hw + bandwidth than prioritizing than simply provisioning for peak load of a particular - even important - application.

A large company needs to share infrastructure across the organization already.  BestBuy needs to make sure their POS systems share bandwidth with other applications in the stores, for example.

While QOS isn't guarantee'd across the public internet, it is purchasable between a customer and a cloud infrastructure vendor. AWS, for example, offers this service. 

People like to buy boxes. It makes them feel safe.  Oracle is likes selling boxes.  But is a lot of boxes the best way to provision for peak load or do we just have to make sure QOS is taken care of?  I'm sure there was a day when Ford ran their own copper between their offices.  I bet they don't do that anymore...