Savvis' Project Spirit is a new service offering which could be characterized as a private cloud service, but is more properly a private data center. Project Spirit is an ambitious service from Savvis that abstracts gives companies control over load balancing, switching and routing, server management, and performance analysis tools to decide not only when to scale up or down, but what components need to be scaled up and down. Savvis hopes to have a beta offering by the end of 2009. Savvis has also has a demo video.
Unlike Amazon's Virtual Private Cloud service which is simply a segmented and privately addressed portion of Amazons AWS that is accessed over a VPN, Project Spirit is a virtualized data center with most of the functions you'd expect to have in a data center under your control. Savvis is working closely with Cisco to deploy the Nexus 5000 and 7000 data center switches and the Nexus 1000v virtual switch on VMware vSphere installations. Project Spirit maybe one of the early instantiations of a converged storage and data network. Savvis offers multiple access methods to the virtual data center including MPLS VPN, Carrier Ethernet and Internet access.
Savvis plans to implement security functions like firewall and switch/router ACL's, IDS/IDP functions and file integrity monitoring built into the platform and make those options available to end-users. On vSphere, Savvis is working with VMWare's vSecure API to inspect and control data within the hypervisor. Unlike many cloud service providers such as Amazon AWS, Savvis will work with customers who require security audits of the service that are required by programs like PCI. The audits could be on-site or virtual depending on what is needed. While Savvis has 28 data centers globally, customers will be able to restrict their virtual data centers to specific regions. For example, some companies might be required by law to restrict the geographic location of their data and processing.
Traffic management functions like policy based load balancing and routing will be under the customer's control, managed through Savvis' Station Portal. For example, companies can select from three profiles and shape traffic in and out of the virtual data center and between nodes. Shaping traffic within the virtual data center ensures that you can prioritize critical aspects of application performance such as latency priority and ensuring bandwidth is available.
The company promises simplified data center deployments by abstracting the management of the infrastructure through their portal. Performance elements such as server, storage, and network utilization will be reported to the Station Portal where administrators can determine which components need to be scaled up or down. For example, increased latency could be due to storage bottlenecks, high CPU utilization or network congestion. Adjusting the wrong element won't have much impact on performance. If the problem is network congestion, a simple change in QoS prioritization may be enough to address the problem without having to add more bandwidth.