Where does an IP-SAN solution fit?

Recent announcements affirming the ratification of the IP storage protocol from the Internet Engineering Task Force (IETF) and SNIA marks a milestone in the realization of IP-based storage solutions. The time is now right for vendors such as Elipsan to bring the benefits of Storage Area Networks (SANs) to a much larger customer base. The opportunity exists to leverage the benefits of many of the new storage technologies that have emerged over the last few years within the infrastructure of an IP-SAN. These benefits include the ability to build industry standard flexible, scalable cost-effective ways of providing remote mirroring, fault tolerance and high availability using standard switches and components, and easier management of shared storage systems--all at a significantly lower administrative burden.

Efficiencies gained through easier remote management, use of existing products and less wasted resources will all lead to a significantly lower total cost of ownership and much higher return on previous investments. The initial purchase price of systems and the associated cost-per-gigabyte will continue to drop significantly through the use of lower cost technologies, such as Ethernet networking and Serial ATA disk drives.

Background to IP-SAN

Until a few years ago, storage was still seen as an integral part of a file server, directly attached using either SCSI or a proprietary protocol. This is referred to as Direct Attached Storage (DAS) and, today, is still by far the most widely used.

The move towards networked storage started in the '80s with the introduction of Network Attached Storage devices and continued later in the '90s when Fibre Channel (FC) emerged as an alternative method of connecting storage. FC could be used in either a loop technology, with up to 126 devices attached to a loop, or in a fabric with switches connecting multiple file servers and storage devices. This removed the direct connection between file servers and storage, allowing storage to be networked with multiple file servers sharing storage devices on a Storage Area Network (SAN).

A SAN is a network that connects file servers and storage devices. A SAN storage device is a block-structured storage device, serving blocks to a file server. What is important about a SAN is that it separates the storage from servers, allowing the storage to be utilized and managed in much more effective ways than is possible with DAS. This allows storage to be consolidated in storage pools and allocated to file servers, as needed.

Instead of a systems manager having to take a server down to deal with a failed disk or a server that has run out of space, with a storage network servers can be provided with virtual disks with built-in redundancy and the ability to grow as the demand for capacity grows. Unlike a Direct Attached environment, storage can be added to a SAN without server downtime and disruption. Separating storage from servers also enables highly available systems with no single point of failure to be developed. If a file server fails, another can easily take over its storage, and if storage is mirrored on the SAN, a file server can easily connect to a mirror of a volume if the main volume (or the path to the main volume) fails.

SANs allow storage-related features such as snapshot and mirroring to be implemented in the SAN fabric, avoiding overhead on file servers and putting the functionality where it belongs on SAN devices, such as storage controllers.

NAS 

In a client-server environment, file servers serve files to clients using a network filing protocol such as CIFS or NFS. A NAS device is simply a specialized file server, which replaces a conventional file server. What makes NAS attractive is the advantages it offers over a conventional file server. NAS devices are appliances, and the most important feature of appliances is their ease of use. NAS appliances are sold as plug-and-play devices, which can be added to a Local Area Network and configured within a few minutes. Configuration is done using simple, user-friendly Web GUIs. When you add a new NAS device to a network, client systems will simply see a new file server with one or more new network shares available.

NAS vendors typically claim that their devices are optimized for file serving. In some cases, particularly for high-end systems, the NAS devices are what are sometimes called "thin servers" running special purpose operating systems designed for optimized file serving. In order to meet the demands of the enterprise, high-end NAS systems have added business continuance features such as RAID storage, enhanced backup options including point-in-time snapshots, local and remote mirroring and high availability. Often, they have a SAN behind them, in which case they may be called NAS heads. However, these features are not specific to NAS, they are really related to the storage devices behind the NAS head. SAN devices can provide them equally well.

The major selling point for low- and mid-range NAS devices has been the fact that they provide an easy way of adding storage to a TCP/IP network. NAS has been very successful as a storage technology, and NAS products have penetrated most sectors of the storage market. At the low-end, very low cost NAS appliances have been sold to small businesses with little technical expertise; mid-range systems with extra functionality such as hardware RAID and mirroring have sold into the middle market; high-end systems with high-performance and business continuance features have been sold into the enterprise sector.

Limitations of Current Network Storage Configurations

In its early days, Fibre Channel (FC) was hampered by interoperability problems. Building and running FC SANs typically required a lot of specialist (and costly) system administrators. Fibre Channel is primarily concerned with transporting block data between servers and storage devices, and it does this very efficiently. However, problems arise as there is no way of controlling access to the storage devices, meaning that every file server on the SAN can access each other's storage. This can lead to one server inadvertently over-writing another's data. FC does offer a solution to this by superimposing forms of access control on the storage devices. If the device is a RAID controller, the controller can check and limit access to individual RAID volumes to particular servers; this is known as LUN-masking. In a Fibre Channel Fabric, there are switches connecting the various nodes together. These switches can also be used to control access to devices, using a technique known as zoning. These techniques work but they are cumbersome to set-up, interoperability can be an issue, and the costs are typically expensive.

A further issue with Fibre Channel is data security. In some computer-room environments, where physical access to the network infrastructure can be controlled, security of data is not really an issue. However, if SANs are extended over a Metropolitan Area Network (MAN) or a Wide Area Network (WAN) then protection of data from unauthorized access becomes important. Currently, the only way this can be implemented in an FC SAN is with additional, expensive and proprietary equipment.

These issues have resulted in FC SANs being an expensive and complex solution, both to install and to maintain--although recently those costs have started to reduce. In fact, a large number of installations actually use Fibre Channel as direct-attached storage, where the performance benefits can still be reaped but the benefits of a pooled storage architecture are not fully realized.

NAS devices do not usually provide business continuance features, except in the expensive high-end products. Low- and mid-range NAS devices may have RAID behind them, but that is usually as far as it goes. If features such as high availability, clustering, remote mirroring or snapshot are required, these can be provided more effectively and with more scalability within a storage area network. SANs are also better when it comes to backup; with NAS, backup must be done over the LAN, whereas if a SAN is available it can be achieved by sending the data directly over the SAN. Modern backup techniques such as Third Party Copy provide a way of performing server-less and LAN-free backup over a SAN and, hence, have a minimal impact on file servers and the LAN.

NAS products suffer a number of other limitations. The first of these is that a file-based protocol is far from ideal to use with a modern high-performance database application. Microsoft has repeatedly warned about the dangers of using a NAS device as the storage for Exchange and SQLserver applications. Another limitation is the fact that NAS boxes create isolated islands of storage. These islands cannot be treated as common storage pools, and tools to manage multiple heterogeneous NAS devices from a single management tool is not technically feasible at the current time (although this may happen someday). Concepts such as provisioning additional storage from one system to another transparently is something the vendors have only just started to grapple with.

The bottom line here is that while NAS and SAN technologies have grown up to solve a number of technical problems with direct-attached storage they, in turn, have not totally solved all of the key issues of storage management--such as backup, reliability, security, and auto-provisioning--at a price point that is realistic for the majority of companies wanting the benefits of network storage.

What Do IP-SANs Offer?

Market surveys and analyst reports say that the benefits of SANs have, until now, been limited to larger companies willing to invest significant time and resources required to install and configure in a new infrastructure. It is clear that Storage Networking technology offers enormous benefits in terms of reduced management costs, easier maintenance and improved systems flexibility. IP-based storage area networks reduce the start-up cost even further while building on a company's existing infrastructure and management skills. Ethernet-based products are ubiquitous, with a cost-per-port that is constantly been driven down by the economies of scale of the market per year.

Benefits of IP-SANs include:

* Much more flexible, scaleable systems to be implemented at a fraction of the additional costs associated with Fibre-Channel, allowing new servers and storage to be added when needed, with minimal disruption to operations.

* Centralized management of networked storage resources using IP-based tools provide much more efficient management of those resources, requiring less dedicated staff time.

* Use of standard-based IP-SAN technology leads to significant increases in functionality, providing additional features such as remote replication, automated storage provisioning, device and storage virtualization, as well as improved backup performance.

* IP-SANs provide much more cost-effective distance-based applications, such as remote mirroring and data sharing using standard Ethernet- and TCP-based connectivity rather than high-cost dark-fiber solutions.

* Security features are built into IP-SAN and existing established security tools--such as VPNs, VLANs, and firewalls--can be readily used in an IP-SAN to boost security levels, as required.

What Technologies Will it Replace/Co-Exist With?

Recent trends in the move from simple Direct Attach Storage (DAS) to more efficient storage area networks and associated improvement in management tools provide some elements of the solution; but for many businesses, building a business case to make a significant investment in new technology--even when ultimately it could provide a cost-effective solution--is difficult. According to IDC, networked storage will account for 67% of disk storage systems by 2005. Even with the market faltering in 2001 and 2002, IDC expects that SAN and NAS will grow at the expense of DAS. NAS revenue is predicted to grow at a 32% compound annual growth rate (CAGR) from 2001 to 2005, while SAN disk systems revenue is expected to grow at a 20.2% CAGR. If existing DAS systems provide a reasonable level of performance today, why the move to networked storage and what advantages does it provide businesses?

IP-SAN vs. FC SANs

So far, Fibre Channel has not brought about the anticipated widespread adoption of SANs, and its use is restricted to more expensive higher end solutions. There remains a large, untapped market sector immediately below the FC market where the benefits of SAN are required but, so far, not exploited. SANs allow storage-related features such as snapshot and mirroring to be implemented in the SAN fabric, avoiding overhead on file servers and putting the functionality where it belongs on SAN devices such as storage controllers. The ability of the iSCSI protocol to scale to the Internet provides a standard way to implement remote mirroring of SAN storage, enabling disaster recovery plans to be implemented. Security in an IP-SAN leverages security protocols developed for the Internet, enabling data to be protected when it traverses public networks.

The benefit of SANs are that they provide flexible, high-performance and highly scalable storage environments. IP-SANs do this using the same low cost, well understood Ethernet infrastructure used by filer servers and NAS devices. This is in contrast to FC SANs, which use much more expensive and specialized components, and with them come higher management costs. This means that IP-SANs can provide a very cost-effective way of providing improved data availability, more effective storage capacity utilization and quicker and cheaper deployment of storage. These factors all result in a lower Total Cost of Ownership.

In the short term, Fibre Channel and IP-SANs will co-exist, with the advent of a number of bridging and converter products. In the end game standard-based technology using Ethernet will win out. Analysts are divided between those who think it will take ten years to those who--rather surprisingly--think it will happen three.

IP-SAN vs. NAS

Both Network Attached Storage and IP-SANs offer end-users tremendous advantages in terms of easier configuration and network management using IP-based protocols. Both NAS and IP-SAN are based on ubiquitous Ethernet networking technology which continues to drive down the pricing of individual components while leaping in performance by a factor of ten every two to three years. Ethernet products are well regarded and users are more comfortable using a technology that is used for all LAN networks.

The main distinction between IP-SAN and NAS is the way in which they transmit data and the management of the storage sub-systems. NAS uses file-based protocols, which is not ideal for database applications, although the integration of NAS and databases still proves popular. Another distinction is that NAS creates multiple isolated islands of storage which cannot be managed as a single resource, much in the same way that Direct Attach operates today. IP-SAN offers the chance to treat all network storage as dynamic reusable resources that can be allocated between servers as simple as setting up a network drive.

What Applications Does IP-SAN Address?

Much discussion in the media has concentrated on the more obscure features of iSCSI and IP-SANs. The reality is that iSCSI will be increasingly used, displacing Direct Attached Storage architectures as companies increasingly come to understand the cost benefits of scalable storage and come to realize the benefits of SAN technology and IP-based storage in general.

iSCSI can be used in my environment where block-based access to the data is desirable. This includes a variety of database/RDMS and file server applications such as SQLserver and Exchange. IP-SAN solutions for a wide variety of back-office and database applications are already starting to appear. A list of the sample applications and systems that are using IP-SAN technology would include:

* Exchange

* SQL Server 

* CRM systems

* ERP 

* Rich-Media

* Video Editing.

Indeed many of the easy implementations will be based around small-scale SAN architectures for:

* Storage consolidation

* Faster and easier back-up (to tape and disk-to-disk)

* Disk-less server architectures.

Alongside these systems, new initiatives based around more cost-effective remote applications will start to emerge:

* Remote mirroring/replication

* Remote backup

* Disaster recovery.

Consolidation Using IP-SANs is the Way Forward

Reports from KPMG, SSB (Salomon Smith Barney) and SNIA amongst others, show that consolidation of storage in a storage network permits multiple hosts to share resources, providing advantages in terms of manageability, flexibility, scalability and lowered total cost of ownership. Easier management comes as result of having one tool that can configure and monitor all the storage in the network; provisioning storage between server and applications can be done at the click of a button, allocating necessary storage resource to match the application storage requirements.

Disk utilization rates in a shared consolidated environment can rise to mainframe levels of 80-85% compared to more normal rates on a DAS environment of 30-40%. This means that by moving to a SAN environment, managers can defer purchasing decisions while they re-use existing space.

Consolidation provides numerous benefits--not only better utilization of resources but also the ability to change environments without significant loss of service. SBB lists the following associated benefits:

* Reduced server hardware requirements

* Increased systems reliability with easier failover architectures

* Labor cost savings

* Faster testing and development

* Easier backup of critical and non-critical data.

Cost savings depend on individual sites and location details but are generally applicable to a wide range of application and production environments.

Comparing Costs of FC SANs and Ethernet-Based SANs

The benefits of SANs have, until now, been limited to larger companies willing to invest significant time and resources required to install and configure in a new infrastructure. It is clear that Storage Networking technology offers enormous benefits in terms of reduced management costs, easier maintenance and improved systems flexibility. IP-based SANs reduce the start-up cost even further while building on a company's existing infrastructure and management skills.

Ethernet-based products are ubiquitous, with a cost-per-port that is constantly being driven down by the sheer economies of scale of the market per year. Compare the cost of a 16-port Gigabit Ethernet switch with an equivalent Fibre-Channel switches ($1,200 compared to $16,000) or the cost of an iSCSI HBA compared to a 1-Gbit Fibre-Channel HBA ($4.50 compared to $1,200).

One analysis shows that the cost to install and maintain a mixed application environment with multiple NT severs and 2 terabytes (TB) of storage in a DAS environment would cost $404,049 over 3 years; the equivalent cost for a FC-SAN system with 2TB shared between servers would be $358,958 over 3 years. The cost for an IP-SAN system with 2TB shared between servers would be $216,693 over 3 years.

Cost-Effective Self-Healing IP-SANs

Elipsan's vision is to provide tools and products to simplify use and management of network storage. To enable this, we are developing comprehensive easy-to-use SAN management software, utilizing existing technologies and products where available and integrating them within an extensible management framework. Key to the process is the development of a number of tools that automate common tasks, such as increasing the amount of space allocated to a particular application with the goal of providing self-managed storage.

Elipsan's IP-SAN management software is designed to provide integrated management of the components of a low-cost, fully featured IP-SAN.

Those components include the storage subsystem of Windows file servers, iSCSI HBAs and initiators, and iSCSI storage appliances.

Elipsan's design goal is to provide a simple interface enabling users to configure the virtualized storage and its various features. The virtualization comes from the comprehensive feature set of their storage appliances, integrated with and exploiting the standard storage management features provided by the Microsoft Windows servers, such as Volume shadowing and Volume snapshot.

Elipsan presents a 'server-centric' view of the IP-SAN, allowing administrators to concentrate on provisioning storage for their file servers. For example, if an administrator needs to create a new 100-giga-byte RAID-5 storage volume for a particular server, he would simply select the file server from the Elantra Web-based management console, and enter his requirements. Elipsan agents would then do the hard work of configuring the storage, such as:

* Search the IP-SAN for storage controllers with free RAID-5 capacity;

* Allocate new iSCSI target volumes on those controllers;

* Set up security on the target volumes so that only the file server specified can access them;

* Initiate 'discovery' by the iSCSI subsystem on the file server, so that it knows about the new storage;

* Run the Windows disk manager and aggregate the storage created into a single Windows 'virtual disk;'

* Create a file system on it;

* Set up sharing of the file system, making it available for use by clients of the file server.

Obviously, much more complex tasks--such as integrating snapshots with the scheduling of back-ups and configuring remote mirrors--are supported. Management agents are integrated with new storage serves being provided under Microsoft Windows 2003, for example, to handle fallow of storage.

Elipsan's first product, the iCS2100 released in August 2002, is an iSCSI-enabled disk array that provides flexible virtualized storage. It is interoperable with leading iSCSI HBAs and initiators, and so allows a low-cost IP-SAN to be deployed. It is provided with a simple, appliance-like Web GUI for configuration and management, and has been qualified with Windows and Linux file servers running a variety of applications including databases and Microsoft Exchange.

Elipsan's next generation of products will include much more powerful IP-SAN management tools coupled with cost-effective, high-density storage platforms The storage controllers will support hardware RAID and volume replication through the use of local and remote mirroring software based on the iSCSI protocol. Comprehensive data security and zoning will be provided, using standard protocols such as IPSec, CHAP, SRP and iSNS. Volume snapshots are also provided to support and complement backup. The management tools provide easy to use, managed IP-SANs, with integrated management of file and application servers and storage controllers.

Future releases will focus on increasingly automating storage management tasks, reducing the burden and cost of storage administration. This will be coupled with increasingly sophisticated data availability features based on open standards, which will provide for high levels of business continuance at a price affordable to the middle market.

Will Parker is product marking manager at Elipsan (Bristol, UK.)

www.elipsan.com