For the past half year at least, Google Search and often all of  Google services have been “out” – just not available while other search engine and Web services such as Yahoo, Bing, Ask, Yahoo Mail and others are up and running. This has implications for Cloud Computing adoption. It also has to be of concern for potential organizations committing to  Google Apps and other Google Cloud Services. Lets look at the outages problem first.

The Google Outages

The following is the most recent outage on July 4th 2010 Sunday:

This was the 5th time and 3rd browser [Firefox, IE, Chrome] – all failed to connect to Google.

So then I tried to connect to Yahoo Search – same search term a minute later:

Yahoo search in Firefox succeeded; so did Bing search in IE.

Finally I try a minute later to connect to Google again:

Still no connect with Google – in about ten minutes Google finally responds.

Now if this was an isolated incident – no problem. But it isn’t. This has occurred at least a half a dozen times in the past month. Keep an Open Eye plans to formally log these events, for Google or any other web service provider, and have written a small MySQL database app to do so.

Here is the reasoning on what is behind these  Google ” outages”:
1)They are not Google outages, but rather my broadband service provider is dropping a critical connection to Google from time to time.
2)They are not Google outages but rather some contracted distributed data service provider that links Google’s millions of machines to key net ports – and they have gone down. This is still serious business depending on how frequently and how long those outages occur.
3)They are Google outages and the failover system is just not being responsive.

The Implications for Cloud Computing

In a nutshell, if Google cannot deliver reliable Cloud Computing services – who can ?

Now Keep an Open Eye believes Cloud Computing has 3 major hurdles to wide acceptance:
1)The risk of Black Swan events incurring major outages in national networks. Not necessarily crippling everybody but large regional services for days or weeks. A Katrina-like weather disaster including tornadoes and 100 year flooding or a New Madrid Mo, California, or Vancouver -Seattle 8-9 Richter-scale earthquake. Or another perverse terrorist strike on 3-5 mega switching centers. I know Black Swan events a)can’t happen and b)cant cripple a “self healing” Internet. Tell those “cant risks” to a)Gulf Coast who relied on oil companies to provide never failed deep sea oil drilling and b)the 8-15 million unemployed/under-employed after always  efficient and effective financial markets froze up in 2008 with the collapse of Lehman Brothers.

2)The lack of standards in the offerings of the major vendors. It is an Okay Corral-like shoot out among the major service providers as to what and how Cloud Computing will be delivered and charged for. Try to get the menu straight for backup, recovery, security provisions, performance measures, and service level guarantees.

3)Cloud Computing, like mobile/iPad apps, is often a new software development system and paradigm
. It can be non trivial to integrate across home systems and the Cloud; but add the complication that the basic languages and algorithms in the two domains are often quite different. This then begs an all or nothing approach to go to the Cloud – certainly not what many shops want.

So if periodic outages ala Google [remember they are selling Google apps and other Cloud Services themselves] continue; that may be the poison pill for widespread Cloud Computing adoption. Then Cloud Computing will have niche or possibly gradual 10 year rollout as redundancy and failover capabilities become more adept.

5 Responses

  1. Seriously? You’re considering it a connection failure when you’re having a DNS problem? Did you maybe check whether the problem was at your end?

    “Can’t find the server” is not equal to “can’t connect” or “connection timed out”. It means that you had a DNS resolution problem… and Google is not responsible for your ISP’s DNS service.

    1. PJ –

      That is exactly the point I raised in the posting – right at the top is that my provider might be responsible. However, I did a ping with no success and a trace that just bogged down. I await the next opportunity [which should be soon enough] – to trace this down further.

      However I suspect that my provider [and the biggest in Canada]can ill afford to be dropping on a regular basis – and I am not the only one in a community of developers to experience the problem.