ARN

Researchers outline cloud computing obstacles, opportunities

The construction and operation of extremely large-scale, commodity-computer datacenters at low-cost locations was the key necessary enabler of cloud computing, say researchers.
  • Bob Brown (Network World)
  • 16 February, 2009 12:49

UC Berkeley researchers have outlined their view of cloud computing, which they say has great opportunity to exploit unprecedented IT resources if vendors can overcome a litany of obstacles.

The 11 researchers' cloud computing forecast is documented in a paper called "Above the Clouds: A Berkeley View of Cloud Computing." The research group works in the Reliable Adaptive Distributed Systems Laboratory (RAD Lab), a 3-year-old outfit funded by companies such as Google, Microsoft, IBM and Sun.

"We argue that the construction and operation of extremely large-scale, commodity-computer datacenters at low-cost locations was the key necessary enabler of Cloud Computing," they write.

But to take full advantage of the opportunity, vendors need to rethink the way they build their products. Application developers need to ensure their offerings can not only scale up, but scale down quickly to satisfy the needs of customers who resort to cloud computing to meet short-term needs. Developers of applications and infrastructure software also need pay-as-you-go licensing models that conform to the reality of cloud computing. Makers of infrastructure software must also craft products designed to run on virtual machines, a foundation technology in large-scale data centers used by cloud computing suppliers.

On the hardware side, vendors need to think big, the researchers say. Vendors need to be thinking along the idea of building container-scale products, that is, those the size of a dozen or more racks. Energy efficiency and flash memory are among other important considerations for hardware makers.

The paper outlines 10 obstacles to cloud computing:

1. Availability of service

2. Data lock-in

3. Data confidentiality and auditability

4. Data transfer bottlenecks

5. Performance unpredictability

6. Scalable storage

7. Bugs in large distributed systems

8. Scaling quickly

9. Reputation fate sharing

10. Software licensing

The Berkeley team also ticks off a bunch of suggested ways to address these obstacles. For example, it suggests that customers go with multiple cloud providers to ensure service availability and calls on the providers to ensure availability by exploiting their massive amounts of bandwidth to thwart DDoS attacks. The researchers call on companies in the cloud computing market to help customers avoid vendor lock-in by getting together on a standard set of APIs that application companies can build to. The researchers argue for use of encryption to protect data as part of a broad security initiative.

The paper also serves to define different types of cloud computing providers, ranging from bare-bones offerings such as Amazon EC2 at one end of the spectrum to more application-specific services such as Google AppEngine.

Despite the many obstacles cloud providers have in front of them, the Berkeley team clearly sees cloud computing as a big opportunity -- not necessarily a low-margin business.

"The apparently low costs offered to cloud users may still be highly profitable for cloud providers," they conclude.