Two weeks after announcing a business version of its Google App Engine application building and hosting service, Google is acknowledging that the performance of the product’s datastore has been chronically deficient for weeks.
To make up for the recent string of outages, slowdowns and errors, Google is waiving datastore CPU costs retroactively effective to the May 31 bill and until further notice.
The datastore problems, which have rippled out to other App Engine components, have been caused by the platform’s growth, which has outpaced server capacity, Google said in a blog post on Wednesday.
“There are a lot of different reasons for the problems over the last few weeks, but at the root of all of them is ultimately growing pains. Our service has grown 25 percent every two months for the past six months,” the blog post reads.
Jason Spitkoski, co-founder and director of Schedule Bin, a Web application that lets employers create and manage employee work schedules, has been using App Engine for two years, and started seeing performance issues in the past couple of months, with a marked deterioration in the past two weeks.
Because Schedule Bin is still in beta, customers have been understanding about the performance problems, but Spitkoski has been left red-faced in front of prospective customers.
“It is more of an issue when we approach new customers, explain the benefits of using a cloud and then show them the product only to find slow performance and having to improvise our pitch to avoid the awkward silence that looms as the app slowly interacts with the cloud,” he said via e-mail.
Google is scrambling to build up the service’s infrastructure to stamp out the issue, but performance is expected to remain rocky for the next two weeks, Google said.
“I was a bit surprised with how long it took them to address the current performance issues,” Spitkoski said. “For a company with so much data and information, I would have expected them to be more proactive.”
The situation is ironic because Google App Engine is a cloud-based application development and hosting platform created so that developers could focus on building applications without having to worry about garden-variety computing issues, such as server problems.
As with other cloud services, App Engine’s selling point is that the vendor, in this case Google, is better equipped to handle IT infrastructure tasks than most, if not all, potential clients, and thus should be entrusted with handling tasks like hardware provisioning and software maintenance.
In short, the main promise of providers of cloud-based platforms and services is that their customers will not need to worry about IT infrastructure breakdowns, certainly not chronic ones lasting weeks, as is the case here.
The problems come at a particularly bad moment because Google made a big splash at its I/O developer conference in mid-May with the announcement of an App Engine version for corporations.
That new version, called App Engine for Business, is in limited preview, but Google hopes to make it more broadly available later this year.
The regular version of Google App Engine debuted about two years ago and is intended primarily for developers of consumer Web applications.
App Engine for Business, aimed at corporate developers creating in-house applications for their companies, will have features that IT departments typically need, such as an IT administration console, a 99.9 percent uptime service-level guarantee and technical support.
Clearly, Google will need to get the platform’s performance problems under control in order to earn the trust of potential customers in enterprise IT departments for App Engine for Business.
“Cloud providers are held to a higher standard of response time and availability than other vendors, to be sure, but given its resources Google should have been able to anticipate — and plan for — spikes in activity,” said Rebecca Wettemann, a Nucleus Research analyst, via e-mail. “This definitely gives Google a black eye in the enterprise space that it will have to work to erase.”
Al Hilwa, an IDC analyst, views the App Engine problem as emblematic of early “teething pain” from cloud vendors at this early stage of the cloud computing market, not as a long-term inhibitor of the cloud model.
“What App Engine’s troubles point out is that at some point the cloud becomes a real physical entity with actual hardware, thus the laws of physics apply and real-life limitations show up,” he said via e-mail.
“It is a bit ironic in that scale elasticity is being pitched as one of the primary drivers of cloud architectures, but of course there is no such thing as unplanned elasticity. In other words, at some point the infrastructure provider has to deal with pre-provisioning for such elasticity,” Hilwa added.
Setting aside the recent performance problems, Schedule Bin’s Spitkoski has been very happy with App Engine from technical, communication, architectural and pricing perspectives, so he’s not ready to give up on it — yet.
“Schedule Bin has been designed so that we can switch our server platform relatively easily, and that is certainly something I would do if Google’s commitment to their cloud started to wane,” he said.