Cisco Blogs

Cloud Computing and the question of Uptime

- February 25, 2009 - 6 Comments

imageAll the dire predictions for the demise of cloud computing that came with yesterday’s GMail downtime were, to be honest, kinda comical. I mean, after all, its not like my Windows laptop ever needs rebooting or the Exchange server ever goes off-line.However, for cooler heads, it does bring up a good question: what are reasonable availability requirements for a cloud based app? Should they be any different (higher or lower) than for an app sitting on a server in your data center.Last week, I talked about the need for cloud service providers to cultivate trust as part of a successful business model, and setting and meeting expectations around availability is certainly part of that.So, what would you want to hear from a cloud service provider that would make you feel comfortable handing over some application workload?I think the simplistic answer is 100% uptime or”5 9s”. I think cell phones long ago showed that the decision making process is a bit more complicated than that: cell phones regularly drop calls and have mediocre voice quality compared to landlines, but they give us convenience, so for most consumers, the trade-off is worth it.Sure you might have apps that need near 100% uptime, but if you need that level or control, and you have that level of utilization, as I mentioned last week, those apps are probably not good candidates for the cloud anyway. So, what kind of trade-off (lower cost, increased functionality) are you willing to accept in return for lowered application availability?

In an effort to keep conversations fresh, Cisco Blogs closes comments after 60 days. Please visit the Cisco Blogs hub page for the latest content.


  1. Benson:completely agree with: we will see a spectrum of price points with an proportional number of 9's. :)As far as the operational excellence, I think that will lead to business benefit, after all, all those operational oops"" translate to increased costs which directly impact the bottom line. Less ""oops"" are going to reduce your cost structure and give you the ability to price more aggressively against less operationally astute competitors and/or b) command a premium for a tangibly superior offering.Omar"

  2. Marc:Agreed--I was thinking more from a cost-effectiveness for systems that are consistently driven at high utilization. James U. did a nice post a while back ( about the cost-effectiveness (or lack thereof).I think for system that need high availability, but have more traditional peaky utilization, a solid service provider could be as good if not better than a traditional approach.Omar

  3. what are the reasons to happen flopping in l3 switch? I was connected my fastethernet port to another switch gigaethernet port but both are l3 swithes. and i modified fastether port speed, duplex, every thing are same for both ports but it is showing flooping in my switch my one is fastethernet port switch. I kindly asking u to help me.

  4. Honestly, I think there is room for a wide range of provider SLAs because there is a wide range of applications. Some are critical to business, others not so much. And if a user can trade a couple ines"" in exchange for lower cost and additional features that might be the right choice for non-critical apps.The other side of the coin is that (some) providers can be expected to operate infrastructure *better* than the typical IT shop. If a provider manages several magnitudes more infrastructure and focuses on IT as a core business then one might consider trusting them to do a better job than internal IT. Perhaps they could even provide higher availability for a lower cost than doing it yourself.I suspect that many folks believe Google is excellent at managing IT infrastructure, which is why a GMail outage stands out. I personally hope that people believe this about other cloud providers, too, including my employer (Savvis) and others. Honestly, over time I *expect* that excellence at managing IT infrastructure becomes the baseline for cloud providers.Cheers,-Benson"

  5. Actually, for certain types of highly burst-able apps (e.g. public portals, consumer video, accounting systems at end of quarter, etc), clouds may provide better availability in peak periods. Think IRS servers on April 15"".Especially in today's increasingly capex constrained environment.I think the question could be evolved to ""what types of apps could/should be 100% hosted in the cloud"" vs. just ""overflow protected""..."

  6. Not sure I agree with you about high availability apps not being good candidates for the cloud. Generic cloud - yes I gree, but an industrial strength utility service provider can probably do it. This is particularly true for externally facing apps such as portals where connections to customers are more imporatnt than connections to internal developers and IT staff. The selection of the Obama administration to use Terremark's Enterprise Cloud for is an excellent example.