One of the concerns cited by people who believe in traditional ways of computing is the issue of service availability in the public clouds. For reasons known only to psychiatrists, they associate availability to the presence of the software inside their organizational boundaries. If we talk to enterprise users who use email system hosted on-premise, we can hear horror stories of how their email was not available at some crucial juncture. Let us face it folks, downtime is a reality for any service that is available over a network irrespective of whether it is hosted on premise or on the public cloud. We should only worry about reducing the downtime to the minimum.
Any organization using on-premise email will face downtime. All they can do is to minimize such downtown but it involves lots of investment both in terms of infrastructure and personnel. Not every organization can afford this kind of investment and, even if they can, they will be wasting quite a bit of them. In reality, very few organizations achieve the level of SLAs they expect from the public cloud providers. In fact, over the years public cloud providers have improved their uptime so much that their downtimes are significantly less than almost all the organizations that use on-premise email.
Yesterday, Google announced that they have significantly reduced the Gmail downtime, setting a standard for public cloud services. Gmail was up 99.984 percent of time which means seven minutes of downtime per month over last year. In fact, this includes accumulation of small delays of few seconds which some of the users experienced and most of Gmail users didn’t even experience this downtime. However, it is a different story with on-premise email systems. According to Radicati Group, on-premise email had on an average 3.8 hours of downtime per month, making Gmail 32 times more reliable than the average email system and 46 times more than Microsoft Exchange on-premise.
Actually, Google is so confident of their architecture and its reliability that they are taking off the terms related to scheduled maintenance downtime that were earlier part of their SLA. This ensures that any Gmail downtime is counted as an unscheduled downtime, thereby, allowing the users to claim compensation under the SLA. Their confidence can also be seen from the fact that they have removed the 10 minute limit on the intermittent outages. Essentially, Google is saying that any Gmail outage is their mistake and they are willing to compensate their paid customers for it. Whether the compensation is comparable to any loss experienced by their customers is an altogether different debate with many nuances to be considered. It depends on the customers’ ability, based on their business needs, to convince Google to agree to an acceptable amount before they even sign any agreement. In short, smaller customers may get shortchanged while larger enterprises extract some levels of compensation. However, it is an altogether different topic to be considered separately.
It is not just Google but there are other SaaS providers who are confident about their availability too. If we can trust the maturity of telephone networks and stop discussing about the availability issue altogether in the context of business, I think we can also take off the availability issue out of our public cloud discussions too. I agree that there is still work to be done to make cloud services on par with telephone services. However, I also think it is time for us to take it off from our thinking that availability is a biggest detriment to cloud use. If we check the statistics without any beliefs or emotions, we can easily see that it is definitely not an issue. I would suggest the FUD mongers to look out for other issues like security where we can still have some rational discussion based on compliance needs, etc.. But any FUD based on availability is a meaningless exercise.
Related articles
- Destination: Dial Tone — Getting Google Apps to 99.99% (googleenterprise.blogspot.com)
- Google Apps makes a new promise: No downtime (zdnet.com)
- Infrastructure Key to Google’s No-Downtime Guarantee (nytimes.com)
- Google Apps Improves Its SLA – No More Planned Downtime (readwriteweb.com)
- Google Promises No Planned Downtime (informationweek.com)
- Google Commits to 99.99% Uptime for Google Apps (webpronews.com)
That’s excellent! I have always found gmail to be more reliable than any in-house corporate email system that I have used. The only time gmail (in Google Apps) has broken is when the DNS MX records were not working. But a good DNS service would solve that I guess.
Available is one big checkmark to remove from the fear list of cloud computing. But how about data security? I have not seen anybody publicly put out a good story on that. (But I’m sure it’s coming).
Data security is a bit more dicey issue which is also a difficult problem. I am sure with time, we will have better solutions which will take it off from the list.
Availability will still be a concern in areas such as healthcare, finance, and credit card. The more we give availability or uptime a long leash and take it off the cloud concern list, the less cloud vendors will take High Availability seriously as a norm for their cloud architecture. Let’s not lower the bar.
I am definitely not lowering the bar. I am arguing that it is less of a concern. Now that the industry itself if setting the bar too high, the market forces will make sure that the services are highly available.
What you are saying is true only for mobile employees using selected apps. The most vulnerable point of corporate access is “the last mile”, the company’s connection between their building and the internet. For employees in the building using an ERP system, the up time for locally hosted apps can be nearly 100%. For someone writing a letter, creating a spreadsheet, or powerpoint presentation, using Google Docs (or other web hosted app) an interruption in the internet connection means they start twiddling their thumbs.
For email, I agree – web hosting is the way to go. In this case, the risk of interruption due to a last mile cut is exactly the same. For other issues, the web hosted solution is better.
Until the internet “always on” condition is, mission critical apps (other than email) are still better off in-house.
Paul
http://www.linkedin.com/in/pdube
You are assuming that the downtime is only related to network availability. It need not be the case.