Running applications in the cloud can in some ways be a faith-based operation. That’s true any time you throw your life onto someone else’s infrastructure. But when it comes to big cloud services, the consensus is that there is a lack of good ways to see how a particular application is doing, or to even know if a snafu is caused by an app or by a glitch somewhere in the network.
Hyperic, a developer of open source infrastructure management software, staked out this issue in June when it introduced CloudStatus, a free hosted service that provides an independent, graphical dashboard view into the “general health” of Amazon Web Services, including Elastic Compute Cloud (EC2) and the Simple Storage Service (S3). The idea, the company says, is to provide a comprehensive, overall measure of availability, throughput and latency.
Today, just a couple weeks after a problem at the Google server farm left some users scratching their heads, Hyperic is announcing CloudStatus monitoring of Google App Engine.
“We can now give people real-time information about the health of App Engine, so that they can have a useful view of availability, performance and latency,” says Jon Travis, principal engineer at Hyperic. “CloudStatus can expose what is going on with Google’s DataStore [distributed storage system] performance, MemCache [memory caching service] and overall network connectivity. We can tell you how quickly you can read and write records, for example, or how long it will take to update your site from App Engine. By reporting on interactions and latency, we let you know how much you can accomplish in a second or two, how long it takes to make a connection. This is all information you need to take into account to design and develop applications to really take advantage of the cloud, in this case Google’s.”
Along with the new hooks for CloudStatus, Hyperic announced a new plug-in for its Hyperic HQ product — which provides the technology foundation for CloudStatus — that will let App Engine users see beyond general indicators and acquire information about their own application running in the cloud. (The free plug-in is available at Hyperic’s site.)
“CloudStatus caters to more of a global need: How is the service behaving overall? Hyperic HQ is really specific to your particular infrastructure and exact usage,” Travis says. “Google provides a limited dashboard. And Amazon has its red-light, green-light graphics. But CloudStatus provides a broader and deeper view. We’re showing real metrics, and historical data, too. And if you want more personalized, tailored information, you can get that with HQ.”
Hyperic counts several large Web sites as customers, including CNET, and large hosting services such as Rackspace, whose Mosso division uses HQ to monitor its own cloud (a likely CloudStatus candidate).
“By providing a tool for performance monitoring, Hyperic has enabled Google App Engine customers to not only identify and remediate problems more quickly, but also to track and report on critical performance data, such as percentage of uptime, Memcache, and network connectivity,” says Steve Brasen, analyst in the systems management practice at Enterprise Management Associates. “These statistics can be used to ensure service-level agreements (SLAs) are met and customer satisfaction is achieved. This data can also be used to compare the reliability of different cloud vendor solutions. Hyperic is providing very useful metrics — certainly better than anything else available today. Dealing with a cloud infrastructure is more complicated than working with a set of servers primarily because of the large number of variables involved. Hyperic appears to have taken all these variables into account when developing its baseline statistics. And I’m not aware of anyone else developing tools to provide this level of insight to cloud operators or users.”
Beside the ability to work with more clouds, expect Hyperic to add more metrics, more-capable plug-ins and more automation features as for the core product, HQ. Also, Travis says, there needs to be ways for developers to see what’s going on with their APIs. “It’s not just about a service’s availability,” he says, “it’s about the APIs you have to connect to those services. We hear about this from users of EC2 and App Engine. If your APIs are not working, it doesn’t matter how well the service is running.”
Javier Soltero, Hyperic co-founder and CEO, describes CloudStatus evolving into a global monitor-and-manage tool like that running on “regular” networks, with application-specific feedback, and features such as “the ability to get an e-mail alert if something you have running in the cloud is not meeting service levels or in response to some other performance requirement.” HQ’s extensible design will allow addition of new agents that can collect status information from a local datacenter or a cloud. This will make CloudStatus a good tool for companies operating in a hybrid environment, or wanting to watch performance as they migrate applications to the cloud, he says.
So, how do cloud providers feel about someone else broadcasting their performance to the world?
“They’re nervous,” Travis says, “but they realize it’s something that they need, it’s something that’s inevitable. If they can’t provide visibility into the health of their infrastructure, they’re going to have a hard time convincing people that they should pay for the service.” The point, as Soltero said in an interview after CloudStatus’s debut, is not to point out failings or mistakes, but to create transparency and, ultimately, faith in cloud computing.
“The success of cloud implementations as a service is primarily dependent on customer confidence in the reliability of the infrastructure,” Brasen says. “Customers are essentially outsourcing a business datacenter. Organizations are apprehensive about putting business-critical applications and services on an infrastructure that is outside their control. The solution to this is developing SLAs that hold the cloud vendor fiscally responsible for any outages or performance failures that affect business, but such an SLA is useless unless the environment can be monitored to ensure compliance. CloudStatus is really a first step to making cloud as a service truly reliable from a business perspective.”