Cloud storage vendors need to develop long-term strategy to ensure cost

Because of the use of inaccurate billing labels, cloud storage service providers or users and their own losses. The researchers suggest using the hard drive as a charge.

A researcher at the American Association of Advanced Computer Systems (USENIX) said that the use of inaccurate billing standards, cloud storage service providers or to users and themselves caused losses.

Existing cloud storage Services billing method Standard: Storage data volume vs.

"In fact, the most expensive place to use is storage rather than input/output or bytes," Matthewwachs, a researcher at Carnegie Mellon University, said at a recent Usenixhotcloud seminar in the Polish city of the United States. Therefore, the cloud storage system should be based on the use of hard disk billing. ”

Wachs and other researchers at Carnegie Mellon University and VMware surveyed the theme of "Exertion-based Billing for cloud storage rights" in Usenix's research report.

The report said, "Cloud storage authority billing should be exertion-based, to the service lessee actually take due to its input/output behavior caused by the cost rather than the cost of these inaccurate proxy mode (such as bytes or input/output quantity)." ”

Currently, the cloud storage provider of IAAS (infrastructure, services), including Amazon and Google, is typically charged with two factors, namely, the amount of data stored and the number of input/output data on the cloud.

Wachs that charging based on the amount of stored data is a reasonable standard, and there are some problems with the amount of input/output data, because it needs to read the hard drive to transmit and transmit the data in. The cost of transferring hard disk data can be very different each time.

"As a result, the cost of a service tenant acquiring a cloud storage service may be very little associated with the actual cost. ”

Random Access vs sequential access: Questioning the rationality of charging according to data transmission volume

Wachs lists a number of possible factors, the biggest difference being random access and sequential storage.

For sequential storage, part of the data in the hard disk is written or read in a continuous stream of data. In random access, the head needs to look around for different parts of the hard drive to read.

Wachs says the two different ways of working will bring a very big gap.

For example, sequential storage can achieve 63.5MB of throughput per second on a regular hard disk, while random access can only reach 1.5MB per second.

In fact, this difference means that even accessing the same data, users with random access use more system resources than users who use sequential storage.

In the long run, this billing method does not motivate customers to build more efficient data transmission methods and is financially detrimental to users who use this billing approach. It will also erode the profitability of storage vendors, who may not have taken these inefficient actions into account in their initial plans.