Zhong Yun Conforming (3): Qualitative and quantitative research on cloud computing usability

"China Cloud Net Exclusive" Author: Chen Whilin chief advisor in Cloud Network

Summary:

Cloud computing is being used by more and more individuals and businesses, but there is growing concern about cloud computing services in terms of security, reliability, and service-response certainty. While cloud service providers (Clouds Service Provider) generally commit to the availability (availability) scope of SLAs (service level agreement), many cloud tenants do not understand the inherent complexity of usability, There is therefore a lack of ability to assess risk when choosing a cloud platform. This paper first systematically defines and analyzes the algorithm model of cloud computing usability, especially the intrinsic relationship between the IaaS of cloud computing, the availability of PAAs and SaaS. At the end of this article, we have done some statistical investigation and quantitative analysis on the service accidents that AWS has been reported since 2008 to 2012.

1. Cloud Computing Challenges:

Cloud services are being used by more and more enterprises. According to Gartner, the market share of the 2013 public cloud will grow at a rate of 8% from 2012 to $131 billion, as shown in Figure 1.

Figure 1 Public Cloud service market and annual growth rate

In the IaaS (infrastructure as a Service), the growth rate of 47.3%, the market share of 9 billion U.S. dollars. In 2012, IaaS grew by 42.4%. 2016, the public cloud market size will reach 210 billion U.S. dollars, growth rate of 17.7%, and the IaaS will maintain a 41.3% growth rate [1].

However, with a large number of small and medium enterprise CIOs considering migrating their data and apps to cloud computing platforms, the attendant concerns about the quality of service in cloud computing (Quality's).

UCBerkeley, Michael Armbrust of the RAD Lab in Computer science, published a paper on cloud computing Services in February 2009-"adjective clouds:a Berkeley View Cloud". In this paper, the concept model of cloud computing is proposed, and the 10 obstacles that must be overcome by cloud service [2] are Berkeley, as shown in Figure 2.

Fig. 2 Cloud computing model of Berkeley

Of the 10 major hurdles, 1 (availability of Service), 2 (Data confidentiality and auditability), 5 (configured unpredictability), 6 ( Scalable Storage), 7 (Bugs in large-scale distributed BAE), 8 (scaling Quickly) are closely related to cloud computing quality. Berkeley in the explanation of usability (availability), it also specifically mentions the harm of DDoS attacks to cloud computing and the measures needed to prevent them.

In addition, according to the survey data from Newvem, 35% of Amazon's AWS users have basically no defensive measures for downtime; 40% of AWS users do not have regular backups of their data. TeamQuest recently conducted a survey of CIOs in many companies, and 40% of CIOs surveyed said they had cluster downtime when using cloud computing [3].

In the 2012, significant security incidents occurred in many of the best-known public cloud computing data centers. Here are some typical cases [4][5]:

* February 29, 2012 and July 26, Microsoft's Azure incident was 9 hours and 2.5 hours, and many North American and European users were unable to manage and use their company's normal business, and some had completely lost their latest data.

* June 14, 2012, June 29, October 22 and Christmas period December 24, Amazon AWS has a serious cloud service slow and crashes inaccessible problems that affect tenants including many important internet companies such as Netflix, PInterest, Twitter , Instagram et cetera [4]. Each incident causes users to be unable to use the service for up to 9 hours and more.

* July 10, 2012, the famous SaaS (service as a service) company Salesforce Service has a major pause. This is due to the failure of the data center power supply for the company that provides Salesforce company IaaS Service (Equinix). Equinix is said to have regained power in 1 minutes. But Salesforce spent nearly 9 hours to complete the recovery of its related business.

* September 10, 2012, the famous DNS service provider GoDaddy's data center service is paused. GoDaddy management close to 50 million domain names and 5 million Web sites. This service cannot be used normally for up to 7 hours. The reason for this is that the router's data is corrupted. There are also media reports that GoDaddy encountered a powerful DDoS attack. But this claim by GoDaddy denied.

* October 26, 2012, Google's app Engine cloud service paused for up to 4 hours. Google did not post a specific explanation.

* October 26, 2012, the well-known cloud storage provider Dropbox service paused for up to 10 hours. Its specific reasons are unknown.

On the other hand, with the undisputed value of cloud computing itself, many of the service quality problems of cloud computing are becoming increasingly apparent.

Therefore, the qualitative and quantitative analysis of the usability of cloud computing becomes a problem with both research and engineering value. Helps CIOs evaluate a cloud computing platform.

The current academic and industrial sectors have not paid enough attention to the availability of cloud computing, especially public clouds. Lack of qualitative and quantitative work in this area.

This paper first systematically defines and analyzes the algorithm model of cloud computing usability, especially the intrinsic relationship between the IaaS of cloud computing, the availability of PAAs and SaaS. At the end of this article, we have done some statistical investigation and quantitative analysis on the service accidents that AWS has been reported since 2008 to 2012.

2. Cloud computing Availability (Cloud Computing availability)

The availability of cloud computing is a very broad concept. This article defines the availability of cloud computing as follows:

Cloud computing Availability: Includes IaaS, PAAs and SaaS connectivity for all levels of service, reliability, latency, data disclosure and loss, cyber attacks, and any other surprises that cause tenants ' business to be unable to meet expectations, or more serious business suspension. Cloud service providers often use SLAs (Service level agreement) to quantify availability commitments, giving the corresponding availability range of values, such as 99.9% or 99.99, and so on.

According to the classification of cloud computing hierarchy [6], we think that cloud computing's availability (AVAILABILITYCS) includes IaaS availability (Availabilityiaas), PAAs availability ( Availabilitypaas) and SaaS Availability (Availabilitysaas).

We believe that the usability of cloud computing, which is ultimately perceived by users, is closely related to the availability of 3 levels of cloud computing.

In the following section, we first formalize the usability of a cloud computing service and do a corresponding algorithmic discussion. Then, there is a theoretical discussion of the relationship between IaaS, PAAs, and SaaS in the cloud computing layered model.

2.1 Availability

The number of unavailable (unavailable) times that a service takes place in a sampling time range (for example, time t hours) is n. The normal time to run before each is not available is defined as TBFI (times unreported failure). The time that is used to restore the normal operation of the service is defined as Ttri (times to Repair).

Figure 3 Availability of cloud computing services

As shown in Figure 3, the availability of the service within the sampling time T range is:

So we deduced that the availability of cloud services in time t hours was:

which:

MTBFT: In time t, the Cloud service score times unreported Failure[7].

MTTRT: In time t, the cloud service is score to repair[8].

According to Formula 1, we can define a cloud service score time Availability (MTA) under K for a sampling period T:

Suppose a cloud service SLA sampling time t is every day, or say 24 hours. If you examine an MTA for 7,224 hours, the MTA calculation method based on the above formula is:

(editor: Heritage)