Right as the final buzzer was going to sound for last week’s Google I/O conference, I had a really interesting chat with Craig McLuckie, the general manager of the just-launched Google Compute Engine (GCE) infrastructure-as-a-service offering. It’s extremely early days for the platform, and Google’s still working on refining the offering. But here are three things you should know about Compute Engine as you consider signing up for the closed preview.

• Google Compute Engine is best suited for large-scale workloads: There’s a reason Apache Hadoop distribution developer MapR was prominently featured in the first wave of Google Compute Engine partner announcements. While McLuckie does say that GCE will eventually refine its focus and perhaps expand its core competency areas to smaller workloads, in its present state, Google Compute Engine is really targeting the needs of customers who are crunching huge amounts of data. That’s at least partially because Google Compute Engine is running on the same infrastructure that delivers Google’s web-scale production services – infrastructure that’s naturally better-suited to large-scale projects. It’s also why Opscode and Puppet Labs were in that initial roster, as Google moves to help customers leverage DevOps methodologies at scale.

Google Compute Engine isn’t open source, but it understands open standards:

Yes, Google Compute Engine adds yet another IaaS API to the mix, putting the already-unlikely open source cloud standard ideal even further out of reach. But GCE’s API is open source, it leverages the KVM hypervisor for virtual machines thanks to a Red Hat partnership, and McLuckie confirms that workload portability is a customer demand that his team has been paying close attention to. That said, Google is really spending its time getting the platform ready for general availability, and generally speaking, its leaving the portability side of things to partners like RightScale. Oh, and while McLuckie wouldn’t name names, he said that compatibility with “other” open source cloud APIs is certainly not out of the question once GCE gets off the ground.
• Google Compute Engine is designed with the Google ecosystem in mind: Given that GCE enables users to spin up Ubuntu and CentOS-based Linux virtual machines on demand, there’s nothing stopping any user from deploying, say, VMware Cloud Foundry or Red Hat OpenShift in their virtual infrastructure. But users who opt to leverage the entire Google Cloud Platform will find that Google Compute Engine extends credentials from the VM into big data serviceGoogle BigQuery and platform-as-a-service (among other services) from Google AppEngine, making it seamless to use them all in concert. But as that much-mentioned partner ecosystem grows, it’s going to be interesting to see what customer adoption trends look like.

Google Compute Engine is still largely an unknown quantity, and as McLuckie reminded me a few times, it’s still a ways away from full maturity. But he also says that for those large-scale workloads, it’s definitely enterprise ready.

There’s been much made about whether or not Google Compute Engine is an Amazon Web Services or Microsoft Windows Azure killer, largely thanks to its lower prices, but it’s really going to be up to cloud developers to decide which infrastructure provider commands their loyalty.