With David Vellante
Cloud computing creates a different business relationship with IT, transforming technology from a pay upfront, heavy CAPEX model to a pay-as-you-go, on-demand model; paying more if additional resources are needed and less if fewer resources are consumed. This model can be implemented internally as a cloud, or externally to a public cloud, or as a hybrid of the two. The economies come from scaling and sharing resources between departments and organizations to balance out the unpredictable peaks and troughs of a single user to much smoother demand patterns across a large number of organizations.
This shared model brings many technical and organizational challenges, including how to share, how to secure, how to provision, and how to charge.
How to secure, provision, and charge are vitally important components of a realistic cloud model, but the starting point is how to share resources. The key to this is virtualization of the key data center resources, so that the same resources can be shared across multiple users, and if necessary a single user can scale workloads over multiple resources in multiple locations. The key virtualization elements in the cloud model are:
Server Virtualization
Virtualization is the key to sharing computing resources, for computer (server), storage, and network resources. The resources need to be shared and moved between servers, both locally and across distance. Virtualization on servers is a well-established technology, from the early days of IBM CP67 to today’s offerings from VMware, Hyper-V, Xen, and IBM LPARs. Server virtualization over a distance is now possible with products such as VMware’s vMotion.
Storage Virtualization
Storage virtualization has had a slower adoption rate than server virtualization. The initial attraction was improving efficiency within arrays, with implementations such as NetApp’s WAFL system, 3PAR's InForm OS, Hitachi's USP V, and Compellent's Storage Center. This has allowed much better data sharing within an array by supporting thin provisioning, data reduction, and virtual copying. This technology was extended to solutions that offered storage virtualization across heterogeneous arrays such as IBM’s SVC, Hitachi’s USP V, HP’s SVSP (LSI's SVM) and F5 Network’s ARX platform. This type of virtualization has made it possible to move data dynamically among a few arrays with different performance and cost characteristics and create a tiered storage model.
The biggest challenge for cloud storage, however, is moving it over distance efficiently and safely. An effective cloud model requires tiered storage with complete independence from distance limitations, both for cost and data security reasons. The challenge of distance is that the elapsed times for transfer can be minutes, hours, or even days. The cost and speed of communication lines is coming down, but the size of files and volumes are growing; network technology alone is not going to solve this problem. Within a cloud model, data transfer has to be seamless for the end-user. If something goes wrong during a transfer within storage or the network, it will not be good enough to say “let’s rerun the job”. Recovery has to be seamless. Failure to achieve seamless recovery will probably result in a business catastrophe.
From a cloud model or cloud architectural point-of-view, the only way to achieve this is to integrate storage and networking virtualization.
Storage & Network Virtualization Integration
Wikibon has selected F5 Networks as an example of the model that cloud computing will evolve towards. F5 is the only vendor at the moment that offers and has deployed a true integration of virtualized storage and virtualized network across heterogeneous storage in multiple locations. The solution focuses on file-based storage, which is the fastest growing segment of the storage market.
Figure 1 shows the F5 Network implementation, using the F5 ARX appliance to virtualize storage, and the Big-IP LTM for network virtualization. Figure 1 also shows the integration with the VMware vCenter Server for automation of the overall management and monitoring; automation is not possible without this level of integration. The integration and automation of all the components allows the effective use of vMotion for the migration of file-based storage to create a seamless automated tiered storage environment.
This tight integration allows the optimization across all types of network traffic, both local and remote. In addition the integration is a prerequisite for automation of all types of application, server, and storage migration and the implementation of a tiered storage architecture that can span both private (internal) and cloud resources.
F5 Networks is not the only vendor who has a vision of integrating storage and network virtualization. EMC’s VPLEX announcement and the development of the VCE environment by Cisco, VMware and EMC/NetApp are other examples that aim to cover both block and file storage. At the moment, however, F5 has the most robust and mature networked file-storage virtualization and tiered-storage solution available and deployed. The virtualization brings with it the ability of the end-users to see a single file mapping view, regardless of where the data is stored, and allows additional storage functionality such as thin provisioning and virtual copies to be implemented efficiently and seamlessly.
Action Item: CTOs and Infrastructure 2.0 architects should start to look at the integration of storage and network virtualization as an essential part of private and hybrid cloud computing. To achieve this will require:
- Selection of vendors that have a clear vision to deliver integrated network and storage solutions to create a virtualized tiered storage network,
- Changes in IT organization to remove the current stove-piped views of infrastructure architecture and management,
- Development of trust in lines of business in a shared storage and network infrastructure model both internally and externally,
- Development of trust by the lines of business in IT to deliver IT Infrastructure 2.0 that is cost competitive with best of breed service providers,
- Development of trust by IT that the lines-of-business that will effectively manage devolution of decisions of IT resource provisioning,
- Development of show-back or charge-back systems that will allow an effective “market” in internal and external cloud resources to be created and integrated.
This is a multi-year cultural and technical journey for organizations to create an Infrastructure 2.0 environment that will interface with Cloud services, but the strategic importance is hard to over-emphasize.
Footnotes: