I saw an interesting demonstration this week of how WebSphere Portal v7 can be hosted in a cloud using the Amazon Elastic Compute Cloud (EC2). Currently, there are a few specialized portal-as-a-service providers (like Covisnt) and many ASPs will host a portal for you, but being able to take your own WebSphere Portal servers and move them to a farm is a different idea. The point is to provide more elasticity (quickly respond to increases and decreases in demand), better uptime, and lower entry costs (possibly lower total cost of ownership too, but no guarantees there).
This type of cloud usage is what my Gartner peers call Public-Cloud-Deployable, which is defined as “traditional on-premises portal products deployable in cloud-based computing environments.” (Gartner clients can read “Portals in the Cloud Will Take Five Forms”). They emphasize the importance of considering security, integration, and transparency for cloud-based portal hosting.
There were some reasons that putting V6 in a cloud wasn’t easy. Starting and stopping instances has to be simple and automated. And the capability has to exist to keep admin files like logs in a shared location rather than keeping them on each server. Accordingly, IBM added an “enable-farm-mode” in V7 to address these issues.
Back to the demo. EC2 was used with AutoScale, which would start or take down servers automatically when certain thresholds were met (like CPU usage). IBM said this requires thinking about server instances as commodities – if one of them isn’t working, don’t spend time debugging it. Instead, just kill it and replace it with one that you know works. An efficient idea, although I do like bugs being found and fixed at some point.
Some functionality is lost if you want to put the portal in a farm: there is database or grid-based session persistence and failover only, no distributed cache or EJBs, no coordinated task scheduling, and no cluster-wide admin. But if those limitations are acceptable, your future may be in the clouds.