On December 10 and 11 when Google and Facebook went offline, users must have felt like Tom Hanks in Cast Away , abandoned alone on an island. Although the outage lasted only 20 minutes in the case of Google, and a few hours in the case of Facebook, panic and frustration were palpable on Twitter, with #googledown gaining traction within minutes of the outages. These were not part of any orchestrated ‘anonymous’ attacks, or, as a few witty tweeters put it, symptoms of the apocalypse, or of the ‘world going down’ action due later this December.
Google’s outage led to some of its widely used non-search applications such as mail, chat and other cloud-based services going offline. Facebook’s followed, going down for a few hours. Separate infrastructure glitches caused these outages — a buggy software update in case of Google, and Web address translation problem associated with the DNS (Domain Name System) for Facebook.
Although these glitches were not catastrophic in terms of loss of data, they served Internet users another warning against blindly succumbing to the pious platitudes and wholesale endorsements of cloud computing as the next big wave on the Internet.
The hardware hurdle
Soon after the outage, Google promptly reported on the cause of the problem and the actions that it took, in a report published on the Google Apps dashboard. A bug in a “load balancing” software update caused the temporary outage of Google services, it said.
When more than one server is catering to user requests, which is certainly the case when Google is serving content via thousands of servers from hundreds of locations, there is a necessity to balance the load (user requests) on each of the servers. To expose a single server to bombardments of requests, leaving others lazing would deem to be poor network design. The arbitration mechanism used to balance the load between servers and server farms, using intelligent programmes running on powerful switching computers, is “server load balancing”.
Software updates need to run regularly in order to accommodate modifications in the server infrastructure. Google engineers located the 18-minute service outage to a bug in the software that ran on some Google applications. This glitch had affected Google Mail, Chat, Google Drive and the Chrome browsers. But Google’s kingpin, its search engine, was immune to this problem.
Facebook, however, ran into a more mundane problem that websites often face. Facebook spokesperson cited Domain Name System infrastructure changes that were carried out, to be the reason for the temporary unavailability of the social networking site.
DNS is the telephone directory look-up equivalent for translating textual Web address to numeric server addresses (IP address). It is the first, and the most important, step to allow clients to reach the servers. When this look-up fails, the website remains inaccessible. The DNS server reconfiguration seemed to have affected Facebook on the desktop version; the mobile version escaped the problem.
When users put all their apples on the cloud, as they increasingly do, these infrastructural “inconsistencies” can have a dramatic cascading effect. Although Google, Facebook and most cloud service providers claim that they have built in multiple levels of redundancies into their infrastructure, the latest outage is yet another example of how millions of users’ data remains vulnerable to the vicissitudes of global networks.
Of course, it would be wrong to attribute the outages at Google or Facebook to technical incompetence. But surely they are pointers to the risks in putting everything on the cloud, and not anywhere else.
Security and privacy
Reliability is an additional concern to widespread implementation of the cloud. There is already a raging debate on issues such as security and privacy on the cloud. Critics of the wholesale movement to the cloud argue that the world has not yet reached a stage where cloud computing has become inevitable.
Regular incidents of user accounts being wiped off, or hijacked due to weaknesses in the authentication mechanisms of major cloud service providers are certainly discouraging users to bet solely on the cloud.
Security threats in cloud-based services do not just imply lacunae in the cloud infrastructure, but also the hesitation of service providers in admitting that the security is not foolproof. Without this emphasis, users tend to set easy-to-guess passwords, or common passwords between multiple accounts, aggravating the already-existing security flaws.
In conventional websites, that are not cloud-based, if unauthorised access is gained to a user’s account, the ‘cracker’ in most cases would reach a dead end and be unable to invade other accounts of the user; it would also be virtually impossible for him to reach the user’s local drives. The daisy chaining of accounts on the cloud, however, results in a lower-level of protection.
The cloud’s USP is built on the promise: “Take your data wherever you go.” But the downside risk is that like the domino effect, it can bring down much more when compared to its terra firma counterpart. Most cloud-based storage applications, for instance, allow applications to directly access user data on their hard drive, between multiple devices such as smartphones, tablet computers and desktop computers. Breaking into one of these devices implies access to all other ‘connected’ devices.
By signing on to user agreements, they agree that they will remain continuously tracked, and on logs maintained by the cloud-service providers. Privacy is simply absent on the cloud.
The hype about the “revolutionary” potential of the cloud has been tempered by the realisation that its reliability is still not iron-clad. If anything, the recent outages highlight the need for users to spread their risks across multiple service providers instead of putting everything on a single cloud.
