Scalable server provisioning with HOP-SCOTCH

The problem of provisioning servers in a cluster infrastructure includes the issues of coordinating access and sharing of physical resources, loading servers with the appropriate software images, supporting storage access to users and applications, and providing basic monitoring and control services for those servers. We have developed a system called HOP-SCOTCH that automates the provisioning process for large clusters. Our solution relies on directory services to implement access control. It uses network boot and managed root disks to control the image of each server. We leverage IBM´s global storage architecture to provide storage to users and applications. Finally, interfaces are provided to access the services both programmatically and interactively. We demonstrate the scalable behavior of HOP-SCOTCH by experimenting with a cluster of 40 blade servers. We can provision all 40 servers with a brand new image in under 15 minutes.