The Cisco Global Cloud Index, forecasts that all data center traffic growth is expected to quadruple between 2011 and 2016, from 1.8 zettabytes in 2011 to 6.6 zettabytes in 2016. Cloud traffic will be the fastest growing segment of this growth, burgeoning from 39 percent of data center traffic at 683 exabytes in 2011 to 64 percent of data center traffic at 4.3 exabytes in 2016. In 2011, 30 percent of server workloads were processed in the cloud, and 70 percent were handled in traditional data centers. By 2016, 62 percent of server workloads will be processed in the cloud.
But will cloud traffic growth continue in the face of well publicized service outages? We first called attention to the lack of rapid cloud failure recovery in this piece:
More than 18 months later, cloud computing is still disturbingly vulnerable to outages that can close an online business to be out of service or down for several days, or cause it to lose precious data. ThinkStrategies Jeff Kaplan wrote, “There are inevitably going to be disruptions to service availability, and it’s key for service providers to minimize these occurrences and for cloud consumers to mitigate their risk by having a backup and recovery plan in place and by exploring ways to take advantage of offline service options.”
1. Microsoft Azure: On Feb. 28, a “leap-year bug” caused Microsoft Azure to suffer an extensive, worldwide cloud outage that wasn’t fixed for more than 24 hours.
2. Amazon Web Services: On June 15, an Amazon Web Services power outage cut services to customers for about six hours, affecting its Amazon Elastic Compute Cloud, Amazon Relational Database Service and AWS Elastic Beanstalk, which are run from Amazon’s data centers in Northern Virginia. The Northern Virginia data centers, the company’s oldest and most used, suffered a similar outage in 2011 and another in October, leading some to believe its infrastructure is wearing thin.
3. Microsoft Windows Azure: Azure customers in Western Europe were out of service for ~2.5 hours on July 26, when “a service interruption was triggered by a misconfigured network device that disrupted traffic to one cluster in our West Europe sub-region,” Microsoft said. The company said customers’ storage accounts were not impacted during the Azure outage.
4. Google Talk: Also on July 26, Google Talk chat service used by Google Gmail customers went down for almost five hours. Google apologized when the service was restored, saying in part: “Please rest assured that system reliability is a top priority at Google, and we are making continuous improvements to make our systems better.”
5. GoDaddy: On Sept. 11, web-hosting and email services company GoDaddy said a six-hour outage that disrupted its operations was caused by a networking issue and not by an attack from Anonymous, as the hacker group claimed.
6. Amazon Web Services: Amazon Web Services went down in its Northern Virginia market Oct. 22, causing website outages in an unknown number of companies, including Reddit, Pinterest and Airnb. The outages affected Elastic Beanstock services, followed by announcements of service interruptions with its Management Console for Elastic Beanstock Services, Relational Database Service, ElasticCache, Elastic Compute Cloud and CloudSearch.
Several days later, Amazon wrote a very comprehensive report that a memory leak and a failed monitoring system caused the AWS outage. https://aws.amazon.com/message/680342/
One effected company wrote their website was taken out of service by the AWS outage
7. Google App Engine: Google’s platform for developing and hosting Web applications in Google-managed data centers, went down Oct. 26 for about four hours as it experienced slowness and errors. As a result, 50 percent of requests to the App Engine failed. The company said no application data was lost and application behavior was
restored. Google then said it is bolstering its network service to guard against traffic latency.
8. Tumblr: About the same time as the Google App Engine failure, the microblogging platform and social networking website Tumblr said service had been restored within a few hours and the company promised it would issue a full report.
9. Dropbox: The online storage company also experienced an outage on Oct. 26th. It displayed a message that stated, “Error: Something went wrong. Don’t worry, your files are still safe and the Dropboxers have been notified.”
The Cloud Connundrum:
How can users continue to use cloud computing services, especially for mission critical applications, when failures are occuring with increased frequency and failure recovery takes hours or days?
Read this for corroboration of this assertion: Analyst: AWS Could Lose Customers If Outages Continue
Will cloud computing really be cheaper than using premises data centers? At least one company doesn’t think so. Martin Saunders, product director at Claranet said: “People understand that cloud isn’t cheaper. It is always going to be cheaper for a company to buy a server, stick it in server room and run it themselves. But I think if that’s all you’re focusing on then you’re very much missing the point of what this thing is all about, because it’s a service. It’s not just about buying infrastructure and hardware.”
Saunders said that people often compare the cost of running a server on premise to the cost of running a server in the cloud, and conclude that the cost of the on-premise server can be written off in three years, whereas the cloud
server will be an on-going expense. Meaning OPEX will be higher for cloud then on prem data center!
So will Cisco’s forecast for explosive cloud data growth be realized? I’m certainly not betting on that!