A few weeks ago I was trying to update some files I have stored on a cloud storage service (that will remain nameless). I had moved my files there a while back as a way to make it easier to access them from my various devices and to avoid losing them during the next inevitable hard drive failure. For the most part I’ve been happy with the service, but on this day, I was unable to access the site.
Not good, as I was rushing to make some changes and send the files to a colleague.
Frustrated by my situation, I asked a co-worker to see if he was also having problems. He was, so we did the next logical thing you would expect. We went to the service provider's status page to see what they had to say. According to it, the service was healthy and there were no current service or maintenance notices.
Twitter! Of course. Whenever services like YouTube or Hulu have outages, users light-up Twitter with comments and laments. Sure enough, a quick Twitter search showed that, yes, there was a widespread problem that had started only a few minutes prior, and already there was a trending hashtag.
This example shows what's great about Twitter. It is an immensely powerful platform for creating instant virtual communities sharing information and opinion around a topic of common interest. The Twitter community as a group was able to do a better job than the service provider itself of informing users that there was a problem with the service. I and the other storage service users — at least the ones also on Twitter — had formed an impromptu global network of monitors, watching the service from hundreds of thousands of access points. Together could confirm for each other that there was a service-wide outage.
Well, not really. Yes, I could see a number of people on Twitter reporting that they couldn't access the service, but this was all anecdotal information (along with a fair amount of opinion). I had no idea who these other users were or where they were located. For all I knew we might all be customers of the same internet service provider and maybe the problem was there and not with the storage service itself. In addition, while I could go to Twitter to confirm that I wasn't the only one experiencing an outage -- even as the service provider's status dashboard said everything was okay -- I was still searching for evidence after the fact. There was no practical way for me to be notified proactively, nor was I able to reliably see service performance degrading prior to the outage.
Herein lies the problem for manufacturers, or any organization, looking to leverage SaaS applications -- particularly mission critical email, collaboration, and document storage -- as part of their IT infrastructure. While it may be okay for me to use Twitter to monitor Hulu, you obviously can't operate a business this way. Organizations need the same level of visibility and troubleshooting capability for SaaS apps that they've come to rely on for traditional on-premise applications. This includes:
- Proactive issue detection and alerting
- Quantitative data on application performance
- Ability to accurately measure service level attainment v. target goals
- Ability to identify problem sources so the time to isolate and fix is minimized
That last one is particularly tricky for SaaS since most of the datacenter and network infrastructure is outside organizations' IT perimeters. You can't directly see or touch the server or network equipment and neither can your traditional monitoring and management tools. It's not surprising, then, that we often hear from IT admins that they have had to resort to using Twitter because otherwise they are flying completely blind. It's not enough, but at least it's something.
Despite its shortcomings, there is a lot to be said for the “power of the crowd” that is so fundamental to Twitter. What if we could take that same model and use it to proactively monitor our SaaS applications? First, it would require some type of active monitoring behind your firewall at the locations where users access their SaaS applications. These “sensors” could act like Twitter users, constantly running transactions against the service and collecting data on transaction and network node performance. They would also allow you to proactively detect and notify an IT Admin of any outages or performance anomalies BEFORE they impact your users.
Then, what if we could collect and share real-time performance data from those sensors (yours as well as other users’ sensors) into a global database maintained as part of your cloud service. You’d then be able to access this data to gain visibility into the health of the complete service delivery chain between you and the SaaS provider. For example, you could:
- View current status, alerts, network statistics, and performance trends for one or more of your own sensors to determine if you have service issues affecting a particular location or subnet, so you can point and fix faults in your own infrastructure and get users back online quickly
- Analyze your sensor data with the rest of the crowd to determine whether service issues are systemic to the application provider or the result of downstream internet service provider problems; you may not be able to fix these directly, but with this information you would know which service provider to call and could provide them with details to speed their time to resolution
- Confirm exactly what service levels you are getting from your application service providers, with detailed outage data needed both for internal reporting and for provider service level guaranty refund requests
The goal of every IT shop is to keep their application users online and happy. But with SaaS, that’s more difficult to do because administrators do not have the same visibility that they do with on-premise applications. We, as a community, need to come up with ways to change that. Taking a cue from Twitter, and leveraging the crowd – seems like a great place to start.
Patrick Carey is vice president of product management and marketing for Exoprise, a provider of cloud-based monitoring and enablement solutions for Software-as-a-Service (SaaS) applications. Patrick spends his free time thinking about how companies can get to the cloud faster and stay there longer.