Slashdot Mirror
Amazon and Microsoft Are Running One and Two in Two-Cloud Race (fortune.com)
When it comes to computing capacity for public cloud services, Amazon and Microsoft are dominating the pack. According to research firm Gartner, Google is the third in this cloud race. The conclusion comes as Gartner looks into Magic Quadrant's annual report surveys, which estimates the amount and type of cloud computing services offered for rent by big companies. Fortune reports: Amazon's continued strength will not surprise many considering the resources it has poured into this now-$10-plus billion a year business. AWS "has the largest share of compute capacity in use by paying customers -- many times the aggregate size of all other providers in the market," according to the report. Last year, Gartner's take was that AWS ran more than 10 times the cloud compute capacity as the next 14 cloud players combined. Asked whether that means Amazon's dominance has held steady, grown, or decreased year over year, Gartner managing vice president Rakesh Kumar told Fortune the research firm does not have the exact comparable figure, but that it is "reasonable to assume" that AWS has maintained the same lead this year. The odd man out here appears to be Google, which has been trying hard to win market share from the other two powers and to prove that it is serious about the public cloud market. Google remains the third largest player by Gartner's measures, but it has slipped a bit relative to the top two.
Google didn't pay for a study. I've always been fascinated by the "Magic Quadrants" since the time I was doing some research on firewalls and found 3 of them from the same year, all offering different best options.
I predict cloudness is mostly fad for bigger co's because our current OS's, conventions, and management techniques haven't caught up to the virtualization possibilities that networking offers. Cloud vendors are ahead of the curve because that's their direct goal and job. But the lessons will trickle out to general IT.
When they catch up, orgs will choose to keep most of their hardware on premises for security and a desire to not depend on the financial viability of a cloud vendor.
The OS and/or app design will have to be more divorced from hardware design so that servers or server farm units are easier for internal stuff to swap in and out as needed.
If a server or component of the "internal cloud" gets sick, an app stack could automatically switch to a different hardware unit and/or its spare or replication partner. A new off-the-shelf "cloud box" could then be casually plugged in as needed.
I see no reason why an in-house cloud farm couldn't be ran almost as cheep as a warehouse-style cloud could if the cloud boxes are almost plug-and-play. Servicing 30 plug-and-play box farms shouldn't be significantly more expensive than servicing 30,000. It's only like that now because the app stack has to match the hardware in most current shops.
Small companies are probably still more likely to use external clouds because they don't want to hire server hardware staff, electricians, etc. to build and run on-premises cloud rooms.
Table-ized A.I.