Advanced Computing in the Age of AI | Thursday, May 30, 2024

AWS Rates Highest on Cloud Reliability 

The reliability goal for cloud vendors is "five nines" (99.999 percent). According to cloud outage figures for 2014, a handful of vendors achieved that goal while Amazon Web Services came the closest among the largest cloud providers.

Performance results for 2014 reveal that cloud vendors are becoming more adept at scaling their offerings while maintaining generally high reliability and performance, observers said.

Amazon Elastic Compute Cloud (EC2) registered the highest reliability over the past year with total downtime of just 2.41 hours across all regions as of Jan. 6, according to service tracker CloudHarmony. Of the 20 Amazon outages during the last year, the longest was 1.33 hours in April 2014. Amazon EC2 topped all infrastructure-as-a-service providers with an availability rating of 99.9974 percent, CloudHarmony reported.

Google Compute Engine experienced 72 outages over the last year, resulting in downtime totaling 4.46 hours. That worked out to an availability rating of 99.9815 percent.

Farther down the list, larger public cloud vendors such as Rackspace (35 outages totaling 7.52 hours of down time) and CenturyLink Cloud Servers (276 outages, 26.25 hours of down time) registered results comparable to Google over the last year.

Well down the list of cloud reliability leaders was Microsoft Azure, which experienced a series of service outages that appeared to peak in August 2014. CloudEndure, another tracker of cloud outages and downtime, reported at the time that Microsoft Azure was hit with what it called an “unusually high level of downtime” in early August affecting customers in Japan. The cloud tracker said Microsoft reported “partial performance degradation” beginning on August 8 that affected customers in Japan until the following week.

Microsoft disclosed another “full service interruption” on eastern Japan on August 15 but restored service the same day. CloudEndure reported that affected services included auto-scaling and metrics reporting.

The Japan outage was followed by a cascade of partial service interruptions that first struck customers in the western U.S. then moved on to affect cloud providers in Brazil and, finally, the eastern U.S. throughout the following week.

Meanwhile, CloudHarmony calculates that Microsoft Azure experienced 92 outages over the past year with downtime approaching 40 hours. That translated into an availability rating of 99.9354 percent, among the lowest for public cloud vendors tracked by CloudHarmony.

Amazon EC2's second-half reliability performance tracked what CloudEndure reported was a doubling of Amazon Web Services' performance quality by mid-2014, including 50 percent reduction in performance issues.

Four smaller vendors—IDC Frontier Cloud, KT Ucloud, Speedyrails VPS and Virtacore achieved zero downtime over the last year, according to CloudHarmony.

Google Cloud Storage outpaced Amazon S3 in terms of storage performance, experiencing only eight outages over the last year (14.23 minutes) compared to 23 storage outages for Amazon S3 (2.69 hours), CloudHarmony reported. Zetta Enterprise Cloud Storage followed with 15 outages totaling 4.83 hours followed by Microcroft Azure Object Storage with 141 outages totaling nearly 11 hours of downtime.