Archives ||  About Us ||  Advertise ||  Feedback ||  Subscribe-
-
-
  -  
 
 Home > Cover Story
 Print Friendly Page ||  Email this story
Network Storage-Your network's graduation path

It isn't enough to simply add disks to your servers and store data. How do you archive, manage, protect, and share it with other departments and users in your organization? by Soutiman Das Gupta

Adding disk after disk to your enterprise servers is like putting all your eggs into one basket. A number of troublesome issues will soon crop up. What's the use of having data lying around on servers if it can't be easily accessed? So you need to devise a mechanism by which all the company's locations are able to access the valuable data. Then take care of another thing make sure you don't lose all the data in a disaster. The acquisition of storage needs to be more than just a disk or tape purchase. Your network should graduate to accommodate the storage architecture and function as a whole.

Part of laying a foundation to a storage solution is to understand the kind of OS (Operating System) environment and the technical architecture that defines how storage is connected to the servers which execute the various enterprise applications.

Environments

Storage solutions significantly depend on OS environments. The various flavors of Unix and Windows NT/2000 represent the open systems area. There is a lot of diversity in these systems and many hardware and software solutions may not operate on all of them. The current versions of NetWare, Windows, and all Unix variants have good support for storage integration. IBM dominates the mainframe world with its MVS, OS/390, and z/OS and SSA (Serial Storage Architecture) hard disks. Any solution for such an environment must integrate fully into the IBM platform. Lately, IBM has developed an architecture that integrates with other open environments as well.

Avijit Basu, Marketing Manager, NSSO, at HP India says, "The pre-open system environment or the mainframe era had old, expensive, and proprietary storage technology. It originated at a time when an open system environment was not as demanding in capacity, performance and reliability as it is today."

Chandrasekhar Balasubramanian, Head, Storage Division, Sun Microsystems adds to the above, "In a strategy to control the market some vendors offered costly proprietary solutions. Some of these solutions even have reduced functionalities. But this strategy has changed now that more users demand open network storage environments."

Technical architectures
DAS (Direct Attached Storage) is commonly used for storing data. But it is limited because it does not allow an organization to share data easily. It is not an enterprise network storage architecture in a true sense. It is also inflexible and has short-term cost and technology benefits. NAS (Network Attached Storage) and SAN (Storage Area Network) have evolved as more reliable enterprise network storage architectures. Let's look at all three of them in detail.

'DAS' the way most like it
DAS is the most common form of storage used by companies. In this set up, most computer storage devices like disk drives or JBODs (Just a Bunch of Disks), tape devices, and RAID systems are directly attached to a client computer. They use various adapters and standardized software protocols like SCSI (Small Computer Systems Interface) and FC (Fiber Channel.)

In DAS, I/O (Input/Output) is done in blocks from the server to the storage. This type of storage attachment has set the standard for performance and utilization by the server. With DAS, performance is typically characterized by response time for access to data and by bandwidth for aggregate transfer rate of data.

On the flip side, DAS has limited flexibility and does not allow you to scale outward easily. A company's network will typically continue to grow while data on the various DASs will remain scattered at different locations of the network. This will not allow all the users in the network to access and share the data. Management becomes a hellish task with so many disparate 'islands' of information all over. Data security and protection is equally difficult as no form of centralized control can be established.

It's no secret that companies that are serious about their data cannot rely on DAS as an effective storage strategy. It has too many limitations. And IDC estimates that storage management costs can be reduced by 40 percent and IT administrators can manage 750 percent more storage capacity by moving out of the traditional DAS.

However in a SOHO environment with very few users, a DAS can handle the storage needs fairly well.

NAS-stretching storage
NAS is a specialized file server that can be plugged into the network (LAN) just like a network printer, hence the name 'network-attached'. It provides file-level access to data and uses standardized protocols like NFS (Network File System-Unix-based), CIFS (Common Internet File System-Windows-based), and TCP/IP to communicate. One of the many beautiful features of a NAS is that it can serve both Unix and Windows users seamlessly and share the same data between the different architectures. It is ideal for mid-sized and not-so-large companies who have a fairly large volume of data.

With a NAS, storage does not become an integral part of the server. It has a storage-centric design where the server still handles all the processing of data but the NAS device delivers the data to the user. If you have an overflowing hard disk on your main server, a NAS device can allow you to stretch and offer breathing room. You can move archives or completed projects from the main server to the NAS and still allow users in the network to access it. Data storage, security, and backup management can also be centralized.

Computer systems can access data from a NAS over a network via a file 'redirector' that changes the access to a file from the native file system (on the originating computer system) to a network operation using TCP. The remote NAS device runs software that allows the file system to support an individual client access. The file system on the NAS server determines the location of the data requested by the application client whether it is in its cache or on the storage.

NAS causes overhead for using the LAN since it rides on the TCP/IP protocol stack and consumes processing power. It may bring latency into the network when another processing element is placed in the I/O access path.

Some SAN-ity
A SAN is a high-speed dedicated storage network or subnetwork which can integrate RAID arrays, tape backups, CD-ROM libraries, and JBODs. The SAN network allows data transfers between computers and disks at the same high peripheral channel speed, as if it is directly attached. It is ideal for large companies that have networks across large geographical areas and medium companies who expect quick growth. SANs are used by industries like petroleum, banking & financial institutions, and retail manufacturing.

SAN deployments are largely driven by the use of the FC standard as a common interface. FC makes use of a circuit/packet switched topology capable of providing multiple simultaneous point-to-point connections between devices. It offers advantages like good connectivity over large distances and scalability. The SAN advent is driven by new requirements of applications like data warehousing, data mining, and OLTP (On Line Transaction Processing) which need high bandwidth and tolerate zero latency. The FC interconnection protocol allows data transmission speeds up to 1 GBps, which is much faster than traditional SCSI-based PC and server devices which have a maximum speed of 160 MBps.

SANs see storage as separate from a server like NAS. But unlike NAS, the SAN architecture involves an independent network or subnetwork. It provides its own network to storage and offloads primary network from all storage related I/O and backups. With a SAN, servers are not directly involved in the storage process. They simply monitor it. And by removing I/O from the servers and the LAN/WAN, you free up bandwidth for applications. This allows enhanced network performance and removes traditional bottlenecks. With the use of a SAN switch you can permit concurrent traffic between all servers of the network and share all the storage devices.

SANs support disk mirroring, backup and restore, archival and retrieval of archived data, data migration from one storage device to another, and sharing of data among different servers in a network. SANs can incorporate subnetworks with NAS systems.

FC is without any question the backbone of a SAN architecture but SCSI can be used as the interface to link storage devices to the SAN backbone. This is because FC supports simultaneous transfer of different protocols. SANs also support ESCON (IBM's optical fiber interface).

Instead of putting the storage directly on the network, the SAN concept puts a network in between the storage subsystems and the server. This means that a SAN actually adds network latency to the DAS storage model. SAN standards are still in the formative stage and vendors like EMC, Compaq, and HP have announced proprietary standards. This collection of proprietary architectures may create roadblocks to successful NAS and SAN integration and data sharing between heterogeneous platforms.

The big question
Since there are various storage topologies available, the big question is, how do you decide which is the best for your company? A good way to begin is by studying the differences between DAS, NAS, and SAN. (see table DAS, NAS, and SAN-A brief look)

DAS is appropriate for very low-end applications like home office environments and with portable computers. It can be difficult and expensive to add networking to the storage architecture.

NAS is the only form of storage that optimally supports both NFS and CIFS network file system protocols for sharing storage between Unix and Windows NT hosts. Since it offers standardized, reliable, and integrated file locking, NAS servers are suitable for many applications where business advantage can be gained from sharing data between Unix and Windows NT clients. It is important to note that file sharing applications provide not only business benefits to non-IT departments but also offer a major infrastructure benefit to the IT department like increased storage utilization.

A SAN may not be a good idea if your network has many OSs and platforms. And installation is comparatively difficult. But its advantages are numerous. It easily supports multiple user access without much overhead. The architecture is optimized for high usage, multiple access, and transfer of large files. Storage capacity can scale upward easily using FC switches and storage devices without adding significant costs. Redundant hardware and software components provide high availability and the system can be configured so that there is no single point failure. Such an architecture makes it an ideal choice for large enterprises or enterprises who foresee positive business growth. It can also support a feature called server-less backup.

Owais Khan, Business Manager, Enterprise Storage at Compaq says, "In order to help decide which storage architecture to use, the IT Head needs to make a preliminary review of its current IT infrastructure, data storage capacity, and volume generation. He/she should then map the required storage capacity against the business growth plan. The organization must also consider disaster recovery solutions. Enterprise-wide backup solutions with built-in centralized management functionality need to be deployed. It is also important to consider the levels of service and support that the vendor can provide. Only when certain key checklist criteria have been assessed does the IT Head decide which solution, whether NAS or SAN is more appropriate for the enterprise."

Migration: Turning NAS-ty
"Most users in India use DAS for storage. But as network data rates slowly exceed the capacity and speed of DAS systems, SANs and NAS will certainly present itself as the next option of storage architecture," feels Ramesh Ramnath, Manager, Systems Engineering at Cisco.

It isn't rocket science to migrate to a NAS environment from your legacy server-based storage architecture or DAS architecture. Anand Padmanabhan, Head, Enterprise Products Divisions at Wipro Infotech says, "You can move the existing data into a NAS box and simply plug-it on to the network. This is in simple terms migrating to a NAS."

Sam Thomas, Senior Marketing Manager (Products), Acer India says most NAS solutions support multiple protocols like NFS, CIFS, and HTTP. "So an organization can migrate its data that resides on heterogeneous server platforms to a common NAS server. Servers can then transparently access the data residing on the NAS devices."

Unifying NAS and SAN
In spite of vendor hype and claims to offer integrated NAS and SAN architectures, it is not possible to deploy an optimally integrated NAS and SAN architecture today. There are significant technical challenges that emerge when you try to integrate block-oriented SAN and file-oriented NAS storage access methods.

Blocks are physical data concepts that refer to the organization of data on a disk drive in terms of sectors and tracks. Files are logical data representations which are understood only by a file system and are made of multiple blocks, sectors or tracks.

A disk drive has no knowledge of files. The translation between the logical file data and the physical organization of this data into blocks on a disk drive is the job of the file system. File systems are usually embedded in a client or server's OS (local file system) but remote file systems over a network like NFS or CIFS have become standard for offloading block retrieval processing using redirector software.

"In order to integrate NAS into a SAN fabric the challenge is to connect the NAS storage to a fiber channel switch while retaining its data integrity. This opens access to the NAS storage by all authorized servers within the SAN fabric. This also allows the NAS storage to enjoy the benefits of centralized backup to a tape library configured within the SAN fabric," says Savio Monteiro, Country Sales Manager, Channels of Vertias.

A great deal of NAS storage is SCSI-based, with internal drive and channel architecture that uses SCSI or the latest Ultra3 SCSI. Typically, a SCSI NAS will interface with a 10/100 Ethernet LAN and allow multiple Ethernet ports. Since SCSI NAS cannot directly interface to a fiber switch, the key is to insert an external router that accommodates SCSI input and provides FC output.

Backup
To achieve efficient backup and recovery in an enterprise, you should take a consolidated information infrastructure approach. In a traditional distributed enterprise with decentralized storage, each platform has its own storage and backup process. EMC's Country Manager, T. Srinivasan says, "In order to achieve zero downtime, you have to centralize the information through a networked storage infrastructure. This allows the information to be centrally protected, shared, and managed. The storage software will then enable business continuance on top of the networked storage layer." You can use the usual backup devices like various tape and disk drives.

Disaster recovery
An enterprise can lose data due to a number of reasons like human errors, disk failure, hardware and software malfunction, and natural disasters. There are a variety of hardware and software tools available to perform efficient backup and recovery. Some examples are tape drives, tape autoloaders, tape libraries, and software like Legato Networker, Veritas NetBackup, and CA ArcServe.

R. Prabhakar, Associate VP, Managed Services at Bangalore Labs says, "Traditional means of disaster recovery were to make backups on tapes and storing copies at a remote location. Today there are various options available depending on the restoration time that you expect from the disaster system. Typical restoration times with backups range from an hour or more depending on your data size and backup policy. If you need restoration times in minutes, you need to use disks instead of tapes. NAS and SAN systems allow you to create snapshots of data that can be replicated to remote systems over a traditional TCP/IP network or through dedicated fiber cable."

Whatever may be the choice of storage, most IT Heads need to realize that even in today's tough economy if you create a comprehensive storage strategy, it will translate into competitive advantage and significant financial gain.

Soutiman Das Gupta can be reached at soutimand@networkmagazineindia.com

 
     
- <Back to Top>-  

Copyright 2001: Indian Express Group (Mumbai, India). All rights reserved throughout the world. This entire site is compiled in Mumbai by The Business Publications Division of the Indian Express Group of Newspapers. Site managed by BPD