Client computer storage consolidation.

Managing the storage inside desktops, workstations and servers in an office can be a daunting task as each computer has its own disk drive containing data that must be backed up, restored and archived. Most storage management solutions today consist of management applications running on the client with user intervention enabled. By consolidating the client storage to one central location, organizations can ensure that it is managed by IT professionals, providing high-quality maintenance and the greatest possible uptime.

The Need for Consolidation

Today personal computers fill the office with, typically, one computer per user. Each user creates information and saves it on his local drive. If the user's disk drive fails, the data must be restored from a backup. However, backup programs can have serious shortcomings. In some cases, user intervention is required for tagging important data to be backed up, and the user may not have tagged some or all of the critical work. Also, if the user's data is on a disk drive that is attached to the computer, the only way to get the data will be to fix the computer or to remove the hard drive. These scenarios can result in a significant loss of productivity for the user.

[FIGURE 1 OMITTED]

Finally, because each computer has its own disk drive, the applications and operating system must be installed on each client. Users are responsible for getting updates and patches for the OS and applications installed, leading to inconsistent deployment. Rolling out new and updated applications can also be inconsistent. As a result, there may be many different versions of software among the clients in an organization, making data management more difficult.

Consolidating the client's storage to one location provides an opportunity to alleviate these data management challenges. Backups of the user's data can be performed offline and after hours, even with the client computer turned off. Data is backed up using a consistent policy to ensure that the latest version is restored. Consolidated storage can be protected by the use of a redundant array of independent disks (RAID), with the complexity of RAID hidden from the user, and the system maintained by IT experts.

With RAID in place, the loss of a disk drive does not mean loss of data and subsequent downtime. It simply means replacing the disk and rebuilding the array. If the configuration includes a hot spare disk drive, this activity can take place without the user even noticing. In addition, virus checking of data can be handled offline. The latest virus definitions can be applied before running the scan to ensure that all data is as safe as possible. Another benefit of consolidating client storage is the ability to implement a just-in-time storage capacity policy. Storage usage can be monitored for each client, and capacity can be provisioned as needed for each user instead of devoting large amounts of unused storage space to every client.

Still another advantage of consolidated storage is that the centrally located data can be accessed from other computers. This is beneficial when a user needs to be moved, when a user is unavailable to share files, or when the user's computer is down and needs to be replaced or upgraded. Consolidated storage also means that data can be moved within the central storage location for upgrades of equipment or changes in storage policy (for example, migration from RAID-5 to RAID-6).

Lastly, applications and operating system images can be housed in a single location. Less storage is required because these images are not duplicated for every client. Patches and updates can be installed in one place to ensure that all clients are kept up to date.

Implementing a System for Consolidation

There are three basic elements involved in implementing a system for consolidation, as shown in the Figure. The first element is the client side--the computer to which the end user has access. This must be set up with minimal impact to the user so that productivity is maintained. The second element is the network, which connects the client to the storage. The network should be relatively easy for IT to set up and maintain. Finally, there is the storage backend, which houses all of the user's data. This backend consists of both the disk drives and the management applications.

Client Computer: The client computer will be virtually the same as before implementation, with the exception of not having an internal disk drive. This drive is replaced with an adapter that allows a networked connection to the storage backend. There exist several technologies to accomplish this, including Fibre Channel and iSCSI over Ethernet. Fibre Channel is most commonly used in data centers, and requires specialized equipment and cabling to implement. Today, iSCSI over Ethernet is rapidly gaining momentum because it maximizes reuse of existing infrastructure and leverages Ethernet networking knowledge that the vast majority of IT departments already possess. For these reasons. iSCSI will be used as the implementation example in this article.

The biggest requirement for iSCSI on the client side is the need for a bootable iSCSI initiator host bus adapter. Once installed, the initiator acts as the bridge to the client's storage. and looks like an internal disk drive to the client.

Network: The next piece of the puzzle is the network connecting the clients to the storage. While Fibre Channel requires different cables, switches and connectors than the office LAN, Gigabit Ethernet does not. In most cases, the existing office network wiring infrastructure works fine. Gigabit Ethernet does require Gigabit-compatible switches, but these are relatively inexpensive compared to Fibre Channel, and are typically much more interoperable with each other.

Since IT departments are familiar with Ethernet and TCP/IP, they also know that security measures are necessary to protect data when using these protocols. Fortunately, there are several technologies available today to protect the network from vulnerability. The best mechanism for this particular application may be IP Security (IPSee), which encrypts the data being sent between the client and the storage backend. The sensitivity of the data will dictate if it should be encrypted or not. User authentication protocols and network zoning are two other examples of security mechanisms that are available. Additionally, it may be best to separate storage traffic from regular network traffic. This provides a physical level of protection, and can minimize the reuse of existing network wiring.

Storage Backend

The final system component is the storage backend. Referencing Figure 1, this consists of both the storage (iSCSI Storage) and the management of the storage (on the Servers). The complexity of this subsystem can increase rapidly due to its inherent flexibility. An organization can use iSCSI for the complete delivery system, or iSCSI can be connected to a Fibre Channel backend. The Figure shows an example using iSCSI only.

First, an iSCSI target is needed. The target is an array of disks, which can be configured and divided to provide storage to the clients. A target usually consists of four to sixteen disks. Disk drive types can include IDE, SCSI, and Serial ATA (SATA). The type of disk chosen depends upon system requirements such as uptime, operating life and other features. For example, low cost can make IDE drives attractive, but their lack of enterprise features such as hot plug makes them the least favorable. SATA drives also have a lower price than SCSI drives, and they incorporate some of the enterprise features of SCSI drives at a lower cost. A SATA drive will have a relatively high-speed interface, and some vendors are offering longer time before failure than is typical of IDE drives.

SCSI drives have been the staple for enterprise use for some time. Their high speed, long list of enterprise features, long device lifetimes, and proven track record have made them a popular choice. SCSI may still be the best choice today for those organizations requiring extra dependability; but that extra comes at a higher cost. SCSI is targeted for a replacement technology. Serial Attached SCSI (SAS) will be replacing SCSI in the near future, just as SATA has begun to replace standard Parallel ATA.

The bottom line is that iSCSI target developers are not short on options to choose from when implementing the physical drives. The target may also have one or more connections to the network. It may be best to provide redundant network paths to the storage backend to prevent a network failure from keeping the client out of communication with storage.

Management for the storage backend is very important. To allow for just-in-time storage, an enterprise volume manager program will be needed to manage the volume size. A RAID configuration utility will add protection from disk failures. The use of an iSCSI name server (iSNS) will give the clients the ability to locate their storage on the network. At the present time, all of these tasks are typically performed independently of each other; however, work is currently being done to provide uniform management for storage networking.

Conclusion

Having storage in every user's computer continues to be the standard in the workplace. At the same time, organizations are looking for ways to increase efficiency and reduce total cost of ownership. The benefits of consolidating a client's storage are clear, including easier maintenance and improved uptime. An inexpensive technology such as iSCSI may be the catalyst that allows an organization to change from the current model to one of consolidated storage.

David Griego and Donald Harbin are members of the Storage Components Division, Intel Corporation (Santa Clara, CA)

www.intel.com