Guest Column | April 23, 2009

Storage, The Final Hurdle To Achieving True Server Virtualization


Written by: John Spiers, LeftHand's former founder and now Evangelist

Before you run out and start your server virtualization project, make sure you know what you are getting in to. Virtualization increases the utilization of servers but also demands more from the current infrastructure and can lead to performance bottlenecks. Furthermore, while companies of all sizes can implement virtualization to make high availability and disaster recovery a reality, they need to realize that traditional storage offerings can limit implementations and make it harder to recover from a site failure or disaster.

Virtualization is driving the need for flexible and dynamic storage solutions that can easily adapt to virtual environments. Without that flexibility, storage configurations are difficult to change and when changes are made they are quite disruptive. More importantly, storage management needs to be simple and flexible. Clustered scale-out iSCSI storage is specifically built to leverage the flexibility of Ethernet, making it possible for companies to eliminate management, performance and availability roadblocks as well as achieve the full power of virtualization.

Simple management for storage flexibility
The “set it and forget it” mentality of legacy architecture no longer applies in the virtual world. To meet the requirements of true virtualization, users must be able to make configuration changes “on the fly” such as provisioning volumes and adding capacity in dynamic virtualized environments. With clustered storage, users can adjust storage configurations along with changes made at the application layer without requiring a specialist to handle the task. Users can easily manage remote offices or a “lights out” data center from one centralized location and scale performance and capacity without taking applications offline. Next generation iSCSI clustered storage architectures integrate all features into a centralized management console. This simplifies management and reduces costs dramatically.

Avoid performance bottlenecks that appear when you virtualize servers
With virtualization, end users increase the utilization of application servers, which increases the overall load on their IT infrastructure. In virtual environments, servers are often taken from a rate of 15 percent utilization to 85 percent utilization. The higher utilization rate means the server requires more performance and resources from the network and storage. This often leads to bottlenecks. Performance issues are never easy but become even harder to solve in a virtualized environment. Clustered storage helps avoid these bottlenecks by scaling central processing unit (CPU), controller and cache resources with disk, allowing performance to scale with disk capacity.

Understanding the costs and availability of a new storage system
Two key drivers for virtualization are cost reduction and achieving high availability. While users may experience cost savings from virtualizing their server environment, they could easily spend all of the money they have just saved on storage. Features like thin provisioning become critical to increase capacity utilization and keep the savings achieved with virtualized servers.

For thin provisioning to have a significant impact on capacity utilization, it has to be integrated throughout the architecture. Having thin provisioned volumes, clones, snapshots, and remote copies of data are all important for reducing overall storage costs. Capacity-free volume cloning eliminates having to store multiple copies of the same operating system image or virtual machine files, improving storage resource utilization in virtual environments. Thin provisioned remote copies may be the most important as high availability and disaster recovery is emerging as the top reason why customers are deploying virtualization today. Server virtualization enables new classes of customers to implement application high availability and disaster recovery solutions. These environments require storage systems with the same level of feature/functionality that are also cost effective and easy to manage.

Providing high availability and effective disaster recovery across geographically separated virtualized data centers is very challenging for most customers and has traditionally had a high price tag associated. The old paradigm required expensive remote replication software and a new storage area network (SAN) for the remote site. Prior to iSCSI SANs, a customer would typically have to purchase expensive Fibre Channel SAN replication software and Fibre Channel to IP/Internet bridge equipment. iSCSI SANs based on a scale-out architecture can create virtual, redundant volumes that span multiple sites. What this means is that virtual machines recognize the same volumes, and connect to the volumes through an IP address at each site. This makes it very simple to fail over servers and virtual machines.

For example, when combined with VMware HA, a site failure can be detected within 15 seconds with the virtual machines started up within a minute at the remote site—as opposed to the several hours that are typically required for an administrator to physically go to the site and bring the servers online and connect the virtual machines to a different SAN with distinct volume copies. When failing over to a secondary location the virtual machines are connecting to the exact same SAN volumes that they were connected to at the primary site. In other words, your data is always online, and it’s just a matter of getting your servers and virtual machines re-connected to the always-available SAN volumes. iSCSI storage clusters or pools can be setup for redundancy across racks, floors, buildings and sites many miles away.

Virtualization places new requirements on storage systems
Virtualization is key to enabling a more efficient model of using IT resources and opens up a new range of possibilities for data centers. These include higher server utilization with lower costs, the economies of server consolidation, and HA and DR supported with built-in features. These requirements will further intensify as companies extend virtualization throughout their environments and move toward cloud computing IT models.

Clustered storage benefits include:
  • 50% the cost of traditional SAN offerings for both implementation and operational costs. Affordable SANs help remove the cost barrier of classic SAN architectures and the economic barrier to the adoption of virtualization.
  • Clustered storage removes the limitations of individual physical devices and removes the daily management of physical assets.
  • Work within budget constraints by allowing storage purchases only when needed, not months or years in advance. Allows the addition of capacity and performance with each storage purchase, making the most of storage investments.
  • Provides built-in high availability and disaster recovery features that enhance server virtualization capabilities to deliver better business continuity.
  • Provides management simplicity, enabling server and storage administrators to manage the SAN and make the most of limited administrator resources and without requiring specialists. By removing complexity from the environment, IT staffers are better able to meet the needs of the business and the users. This also allows the business to operate continuously and without disruption.
With the right storage solutions, IT organizations can truly unleash the power of virtualization across the infrastructure to mitigate risk and increase business productivity.

John Spiers is LeftHand's former founder and now Evangelist.