iSCSI advantages and solutions for businesses.

Microsoft Windows and Linux have intrinsic volume management and RAID support that allow for sophisticated storage configurations. When combined with iSCSI, the result is a network storage solution that is powerful, easy to deploy, and very scalable for any business. The combined technologies simplify the management of volumes by allowing the operating system to assume the management role rather than third-party software applications. This shift in responsibility for volume management will help IP-based SANs to proliferate by reducing the cost of deployment. This article highlights the significant business advantages introduced by new iSCSI solutions when used in conjunction with storage management software included with Windows and Linux.

iSCSI

iSCSI is a quick, efficient and cheap method of implementing network storage solutions. The Internet Engineering Task Force (IETF) ratified the iSCSI specification in February of 2003. iSCSI utilizes the TCP/IP protocols to transfer SCSI information over a network. iSCSI is fast gaining popularity due to:

* Low cost that is a fraction of Fibre Channel

* No limitations on distance

* Easy migration and administration since iSCSI leverages existing knowledge of Ethernet and SCSI

* High interoperability due to the leveraging of existing, proven technology

iSCSI accelerators are available today that accelerate processing of iSCSI and TCP/IP packets. The accelerators offload iSCSI and TCP/IP processing from the host CPU, and the result is faster iSCSI packet processing with reduced host CPU utilization.

Dynamic Disks

Dynamic disk is a technology that Microsoft developed to implement volume management and RAID. Microsoft introduced dynamic disk technology in Windows 2000. Here are some basic facts about dynamic disks.

Supported in these Windows operating systems: Windows 2000 Professional (Limited), Windows 2000 Server, Windows 2000 Advanced Server, Windows 2000 Datacenter Server, Windows XP Professional (Limited), Windows 2003 and Windows XP 64-bit. (Note: Windows XP Professional and Windows 2000 Professional do not support disk fault tolerance.)

Windows operating systems that do not support dynamic disk: MS-DOS, Windows 95, Windows 98, Windows ME, Windows NT 4.0, and Windows XP Home.

Due to these operating systems' inability to interpret dynamic disks, they cannot directly read or write data to a dynamic disk or a dynamic volume. Moreover, dual-boot capability is not supported on dynamic disks nor is dynamic disks supported for removable disks or notebook computers regardless of the operating system or version.

Dynamic disk support is at the drive level and not volume level. Either all or no drive volumes on a disk must be converted to dynamic disk support (i.e., all-or-nothing proposition). Dynamic disks can co-exist with basic disks within the same system. (Note: "Basic disks" is the term used by Microsoft to refer to today's disk partitioning and volume scheme.)

[FIGURE 1 OMITTED]

Dynamic volumes can be formatted with one of three file systems: NTFS, FAT16, or FAT32. The NTFS file system is preferred in order to obtain the most flexibility with options, especially security.

Dynamic disks support five volume types:

* Simple: Basic partition with no spanning or fault tolerance. This is the only option available for single disk systems using dynamic disks.

* Spanned: Multiple dynamic disks or multiple dynamic volumes are combined to form one single, large volume.

* Striped: Volume striped across multiple dynamic drives. Basically RAID-0, no fault tolerance.

* Mirrored: One-to-one mirroring of drives. Basically RAID-1, full fault tolerance.

* RAID-5: Fault tolerance using parity.

(Note: All functionalities are completely software-based and require no hardware support.)

Even though Microsoft does not publish the dynamic disk data structures format, the Linux community has developed a Logical Disk Manager (LDM) driver for dynamic disks. The Linux driver is quite mature and stable.

The biggest advantage of dynamic disks is the ability to manage disks and volumes without rebooting the operating system. For example, a volume can be dynamically extended or changed, and the operating system need not be rebooted. This no-reboot policy also applies to repairing mirrored and RAID-5 volumes, unless of course the driver for the drive requires the reboot.

All dynamic disk functionality is intrinsic to software. No third-party application is required.

Analysis of iSCSI and Dynamic Disks

iSCSI with Windows dynamic disk support is a powerful combination. The big advantage iSCSI introduces is the ability to build IP-based SANs (while utilizing existing network management applications that are familiar and well-known) for a much lower total cost of ownership (TCO) than a Fibre Channel SAN. By configuring iSCSI target drives as dynamic disks, one can easily maintain a scalable storage solution without the need for third-party proprietary applications. The combination of iSCSI and dynamic disks has great applicability to IT infrastructure such as databases, e-mail servers, media streaming servers, and other server types that benefit from network storage.

Microsoft Exchange is a classic example of an application that can benefit from the combination of iSCSI and dynamic disks. For Exchange, anticipating drive storage requirements is difficult at best. Over time as an organization grows and drive prices drop, the ability to easily and transparently expand the Exchange data drive is important. For example, if a drive holding Outlook e-mail messages is filling to capacity, there are only a couple of options:

Employ a Fibre Channel SAN. This option is expensive and is difficult if one is already not in place. A (FC) SAN has the advantage of adding storage capacity without rebooting servers (in most cases). Note that NAS is not an option since Microsoft Exchange requires block-level storage (see Microsoft Knowledge Base articles 317173 and 314916).

DAS configurations. This option requires the administrator to replace smaller drives with larger drives. This option is clumsy since the administrator must backup the Exchange data from the old drive and then restore the Exchange data to the new drive. Furthermore, replacing hard drives is a time consuming job, which renders the e-mail system useless for the duration of this task.

With iSCSI and dynamic disks, an improved third option is introduced where dynamic disks are able to pass on the advantages of FC to the iSCSI world. Simply put, dynamic disks used in conjunction with iSCSI allow many of the volume management benefits of Fibre Channel to be readily made available for iSCSI.

Applying the third option to our Exchange example, the Exchange data would reside on an iSCSI target drive. Figure 1 illustrates an e-mail server that could be running Exchange with the iSCSI storage in a separate chassis (the e-mail server and iSCSI targets could be in the same chassis, if desired). In this scenario, the administrator can easily add additional iSCSI target drive(s) and configure them as dynamic disks, then add these drives to the volume where the Microsoft Exchange data resides. Under these conditions, the volume holding the Exchange data would in essence span to the new drives and painlessly increase storage capacity. The noteworthy point here is that the administrator can perform these tasks without rebooting the system. This is important for mission critical applications since the storage expansion for Microsoft Exchange was added without involving the (Exchange) application as the storage expansion occurred transparently to users of Outlook. The same transparency also applies to replacing failed drives in a RAID volume.

iSCSI initiators and targets, as a combination, allow for optimization of performance while enhancing flexibility. As discussed for Exchange servers, iSCSI initiator accelerators that provide full iSCSI and TCP/IP offload can be used to expedite data across the network to iSCSI targets holding the Exchange data. This configuration of Exchange servers provides the most efficient movement of data over existing IT infrastructure, resulting in enhanced performance with high throughput and IOps.

Linux Volume Management

Linux has long had much of these volume and RAID management functionalities using the Logical Volume Manager (LVM) to configure volumes and Disk Druid to configure RAID volumes. Furthermore, Linux also does not require a reboot when volumes are expanded or contracted, which means existing applications can continue to run as volume changes are made.

For example, Apache is the most popular web server used today with over 60% market share; predominantly running on Linux. If the volume that the web server database files reside on were reaching capacity, an iSCSI drive could be added in Linux, and in turn, the drive could be added to the volume for the web server. The addition of the iSCSI drive would, thus, increase the volume storage space. Best of all, the web server and corresponding database engine would be unaware of the volume changes.

Another example is using the snapshot functionality available through the LVM. The LVM has the capacity to create snapshots of volumes, and these snapshots can then be used to backup volumes, even though data on the volumes may be changing. The snapshots are temporary and only exist as long as the backup snapshot is taking place. After the backup, the snapshot can be un-mounted, and the drive space then returned to the Volume Group. iSCSI allows drives to be easily added and removed using the LVM; hence, storage space can be easily added and removed from the Volume Group. This allows storage to be readily available and re-used for backup snapshots. Moreover, the amount of storage for snapshots can easily be increased by adding more iSCSI drives, since iSCSI allows drives in remote locations to quickly be brought online.

Linux volume management tools provide for Linux today much of the volume management tools that have historically been implemented only for FC SANs. Moreover, since new iSCSI accelerators also support iSCSI initiator and target functionality in Linux, the benefits of iSCSI accelerators such as Xiran's, easily transfer to Linux to provide a complete and flexible iSCSI solution.

Over time, volume management and RAID support will be even better integrated into the operating system and become more ubiquitous. Combining this intrinsic support with iSCSI provides powerful storage network solutions for many existing applications for a TCO that is a fraction of the cost of Fibre Channel. The TCO drops even further while increasing iSCSI performance when Xiran's iSCSI+ solutions are utilized. Xiran's iSCSI+ solutions provide:

* A protocol acceleration engine incorporating Xiran's patented DirectPath technology

* High performance with full TCP/IP and iSCSI offloading

* An easily scalable solution to match today's performance requirements

* Simultaneous target/initiator functionality

Xiran's iSCSI+ solutions increase flexibility because they provide iSCSI solutions for both ends of the wire and everything in between. Xiran's iSCSI+ target accelerators utilize Xiran's patented DirectPath Engine (DPE) to efficiently move data from the network to the storage subsystem without involving the host CPU. This allows a single server to accommodate more iSCSI targets than competitor products while simultaneously reducing TCO. Xiran's DirectPath technology employed in the DPE also allows for the expedited movement of Exchange data across networks by providing full iSCSI and TCP/IP offload. Xiran provides this configuration for Exchange servers as a kit, offering the highest degree of flexibility for IP storage of Exchange data. ISCSI, combined with existing volume and RAID management functionality found in today's operating systems, will accelerate the adoption of iSCSI for IP SANs in business IT infrastructures.

Robert Lusinsky is IP Storage Senior Applications Engineer at Xiran (Irvine, CA)

www.xiran.com