Although SAN’s and SAN file systems have been around for more than a decade, many misconceptions about these technologies still exist. To provide clarity, the following background section includes key definitions and introduces the four major computing environments. Additionally, the attribute comparison section provides an in-depth look at how these computing environments stack up.
Local area networks (LANs) are networks that transfer data between computers. Storage area networks (SANs) are networks that transfer data between computers and storage devices. LANs and SANs consist of a series of switches, hubs, and cables. Both operate within localized geographical areas such as houses, office buildings, and campuses. SANs do not replace LANs; most environments with a SAN also include a LAN.
File Systems / Filesystems
A file system is software that organizes data residing on storage devices and presents the data to users/programs as either files or folders/directories. There are three types of file systems: local file systems, LAN file systems, and SAN file systems. With local file systems, sharing files between computers involves manually copying files onto media such as USB drives and CDs or manually copying files over networks using email and programs like FTP. In contrast, LAN and SAN file systems share files via automated network transfers. SAN file systems are also known as cluster file systems, shared file systems, shared storage file systems, shared disk file systems, multi-initiator file systems, distributed file systems, and global file systems.
A storage island represents the isolation of stored files to one or more computers. Computers with automated access to these files form an island. Computers not part of the island require manual efforts to access files. Local file systems limit islands to single computers, whereas LAN and SAN file systems facilitate islands containing multiple computers.
Direct-Attached Storage with Local File Systems
A direct-attached storage (DAS) device is storage hardware housed inside a computer or connected straight to the computer. A DAS device is typically paired with a computer running a local file system and forms a single-computer storage island.Figure 1 illustrates two computers, each running a local file system, connected to DAS devices. The dotted lines surrounding the computers and DAS devices represent storage islands. There is no automated file sharing between islands. Computer A has access to files stored on devices 1 and 2 but does not have access to files on device 3. Conversely, computer B has access to files on device 3 but does not have access to files on devices 1 and 2.
Key Point: Single-computer storage islands only share data by means of manual file transfers.
LAN Storage with LAN File Systems
A LAN storage environment is a storage island containing a file server and one or more storage devices and computers. Users/programs, running on the computers, utilize LAN file systems to retrieve and store files managed by a centralized file server. All file requests pass across the LAN to the file server; in turn, the file server accesses files residing on the storage devices. The file server is the sole gateway to files and is therefore responsible for maintaining data coherency.The file server is either a dedicated computer known as a network-attached storage (NAS) head or a non-dedicated computer capable of performing tasks beyond just file serving. File server storage devices are often DAS devices managed by local file systems.Figure 2 illustrates the components of LAN storage. As signified by the dotted line, these components form a network-wide storage island. A LAN file system, running on the computers and file server, automates LAN file transfers between the computers and the file server. The file server utilizes a local file system to manage files residing on the DAS devices. Files residing on the storage devices are only accessible to computers via the file server.
Key Point: Within a LAN storage island, all file requests must pass through a centralized file server.
SAN-Attached Storage with Local File Systems
By definition, SANs transfer data between computers and SAN-attached storage devices. Single-computer storage islands form when SAN-attached storage devices are paired with computers running local file systems; the local file systems are incapable of ensuring data coherency required for multiple computers to access the same stored data.Figure 3 illustrates two computers and three SAN-attached storage devices connecting to a SAN. These computers utilize local file systems and thus reside within separate storage islands. Computer A only has access to files residing on storage devices 1 and 2, and computer B only has access to files residing on storage device 3.
Key Point: SAN-attached storage devices paired with computers running local file systems form single-computer storage islands that are limited to manual file sharing.
SAN-Attached Storage with SAN File Systems
SAN file systems organize data residing on SAN-attached storage devices and are responsible for maintaining data coherency when multiple computers have access to the same stored data. Unlike local file systems on a SAN, SAN file systems facilitate a network-wide storage island. Every SAN-attached computer has access to every SAN-attached storage device.Figure 4 illustrates a network-wide storage island containing SAN-attached storage devices and computers running DataPlow SAN file systems. Computers A and B share access to files stored on all three storage devices via the SAN.
Key Point: Computers running SAN file systems have access to all SAN-attached storage devices thereby creating a network-wide storage island with fully automated file sharing.
File sharing makes data accessible to users/programs running on different computers. Automated file sharing exists between computers within a storage island. Manual transfers are necessary between islands.DAS and SAN-attached storage devices paired with computers running local file systems form single-computer storage islands and are thus limited to manual file sharing. These two configurations score poorly because manual file sharing is inconvenient, time-consuming, and prone-to-error.
In contrast, LAN storage managed by LAN file systems and SAN-attached storage running DataPlow SFS form network-wide storage islands with fully automated file sharing. Both configurations receive excellent ratings because the automation:
- Simplifies data organization and management
- Promotes collaboration
- Enables the creation of a new asset - a centralized data library
- Facilitates data-driven decision making
Manageability – Load-Balancing and Expanding Capacity
Manageability is the ease at which system administrators fulfill the needs of users. Well-managed environments allow users to easily locate and access data at targeted performance levels. Although manageability has many components, load-balancing and expanding capacity are key.Load-balancing is either a manual or automated relocation of data from one storage device or island to another. Environments with balanced workloads enhance system performance, whereas imbalanced environments lead to poor computing resource utilization and thus poor performance. Datasets and user activity are usually dynamic, changing throughout the day, week, month, and year. With ever changing workloads, the process of manual load-balancing is difficult, if not impossible, to optimize. Besides being arduous, and often ineffective, the task of manually load-balancing data is disruptive to users because the changing locations make it difficult for users to find their data. For example, files found on C: one day might be found on D: the next.Adding computers, file servers, and storage devices are ways of expanding capacity. Storage networking configurations inherently dictate the nature of expansion. Well-executed expansion cost-effectively accommodates short-term and long-term growth without complicating overall management efforts.As the number of computers and amount of stored data increase, DAS configurations quickly become difficult to manage. If a computer is underperforming, manual load-balancing between storage islands becomes necessary. Manual load-balancing is disruptive to users and results in time-consuming, and reactive, problem solving for system administrators. Therefore, DAS rates poorly for load-balancing.DAS configurations also rate poorly for expanding capacity. Every new computer added to a DAS environment forms a new storage island to be managed. Furthermore, to cost-effectively and equitably add storage is difficult because a single computer receives the entire benefit.
Management of LAN storage is fairly easy for environments with only one file server. For environments with multiple file servers, management is difficult because stored data must be manually load-balanced in order to optimize overall system performance. Accordingly, LAN storage configurations rate as fair for load-balancing.Expanding capacity of LAN storage also rates as fair. Adding new computers and additional storage capacity is typically easy. Computers join network-wide storage islands with fully automated file sharing. Storage devices connect to file servers and provide additional, shared storage capacity. Challenges arise, however, when adding file server capacity. Underestimating future workloads leads to the addition of multiple file servers over time, thereby creating more user disruptions, added expense, and an even more complex computing environment to administer. Overestimating future workloads proves expensive.SAN configurations with local file systems rate poorly for load-balancing. Like DAS, load-balancing between storage islands is disruptive to users and time-consuming for system administrators.SANs with local file systems, however, receive an excellent rating for expanding storage capacity. With minimal effort, system administrators are able to assign storage capacity from any SAN-attached storage device to any SAN-attached computer without disrupting users. In addition, advanced features, like RAID and snapshot backups, are easier to cost-justify because every user benefits.
SAN-attached storage configurations with computers running DataPlow SFS receive an excellent rating for both load-balancing and expanding capacity. Every computer running SFS has access to every SAN-attached storage device; ergo manual load-balancing is avoided. New computers and storage devices seamlessly integrate into the SFS environment. Computers join the network-wide storage island with fully automated file sharing. New storage devices provide additional capacity to every user of the network-wide storage island.
High Availability (HA)
Data availability is the degree in which users/programs maintain access to data despite hardware and software failures. Users expect data to be highly available.DAS rates poorly on availability. If the attached computer fails, data inaccessibility results. Recovery is often time-consuming and involves restoring the computer to a functional state or physically moving the storage device to a different, yet compatible, computer.
|High Availability (HA)
The file server of LAN storage is a single point of failure. Some file servers support high-availability (HA) cluster configurations where backup file servers wait in standby mode ready to take over should the primary fail. LAN storage receives a rating of fair; a high degree of availability is achievable, but due to hardware duplications, HA configurations significantly increase file server costs without improving performance and scalability.Availability of SAN-attached storage with local file systems is similar to LAN storage. The computers are single points of failure that, like LAN file servers, can be backed up with duplicate hardware.
SAN-attached storage configurations with computers running DataPlow SFS receive an excellent data availability rating. By design, SFS detects system failures and automatically recovers without loss of availability.
User and System Performance
Well-designed computing environments address both user and system performance requirements. User performance measures how quickly a computing environment responds to the requests of a single user. In turn, system performance is the aggregate speed of the entire computing environment. Although many factors affect user and system performance, protocol translations and bottlenecks are the primary differentiators.Protocol translations greatly impact user performance. Protocols are communication languages used by computers, storage devices, and file servers. File systems must translate from one protocol to another before initiating data transfers between hardware components. Every file system has a unique translation method. Therefore, every file system generates different translations which in turn produce different data transfer rates which yield vastly different user performance results.Bottlenecks limit system performance. Hardware bottlenecks are computers, storage devices, and file servers that restrict the flow of data. Bottlenecks also occur when users manually transfer data between storage islands. Eliminating these bottlenecks increases system performance.DAS and SAN-attached storage, paired with local file systems, achieve excellent ratings for user performance. Local file systems carry out a single protocol translation for data transfers between computers and storage devices. Most local file systems produce high-quality translations which lead to excellent data transfer rates and thus excellent user performance.For local file system environments, storage devices, computers, and manual data transfers dictate system performance. Upgrading and adding storage devices are fairly simple yet cost-effective methods of eliminating bottlenecks and improving performance. Doubling the number of storage devices, for example, nearly doubles system performance.In contrast, overcoming computer bottlenecks proves far more complex. Adding new computers eliminates processing bottlenecks but, in turn, creates manual data sharing bottlenecks. Computers running local file systems are isolated storage islands which require users to share data via manual transfers. In order to eliminate these devastating bottlenecks, system administrators would need to reconfigure the entire computing environment. Therefore, DAS and SAN environments running local file systems rate poorly for system performance.
LAN storage configurations rate poorly for user performance. LAN file systems must conduct two protocol translations per request. The first translation occurs when transferring data between computers and the file server; the second translation takes place when transferring data between the file server and storage devices. The second translation accentuates deficiencies created in the first. As a result, user performance is quite slow.Storage devices and file servers limit system performance of LAN storage environments. Although storage device bottlenecks are fairly easy to eliminate, removing file server bottlenecks proves far more difficult. Upgrading a file server is straightforward. However, adding a file server leads to load-balancing issues. System administrators are then burdened with the time-consuming and often ineffective task of manually balancing workloads between the file servers. In the end, system performance suffers; thus LAN environments rate as fair.
SAN-attached storage configurations, running DataPlow SFS, receive an excellent rating for both user and system performance. Like local file systems, SFS environments carry out a single, high-quality protocol translation which results in excellent user performance. Storage devices may eventually limit system performance; however, upgrading and adding storage devices are effective methods of increasing system performance.
"SFS paid for itself in recovered disk space."
- Jeff Scott, Los Alamos National Laboratory
"Your cost-effective and robust file sharing solutions are truly industry leading."
- Chris Dobbins, University Center Rochester
"DataPlow SFS is the best solution for server virtualization."
- Max Trinidad, Claire’s Stores Inc.
"Your products and service are superb!"
- Jim McKelvey, VideoCubed
"As a reseller, customer satisfaction is key. I highly recommend DataPlow."
- Jeff Syversen, TDX Tech
"Our business hinges on sharing data quickly. DataPlow SFS delivers with unprecedented performance."
- Greg Kelley, Vestige, Ltd.
"The show must go on, flawlessly. With DataPlow SFS,
I know it will."
I know it will."
- Ross Anderson, Guthrie Theater
"DataPlow SFS is the only software I trust to manage our mission-critical systems."
- Pedro Vivó, Tetrabit
"After implementing DataPlow SFS, our productivity and profits skyrocketed."
- Kevin Fitzgerald, Aero-Graphics
"For web serving, DataPlow SFS is simply the best."
- Jeremy Farber, Streetfire
"I’m pleased with DataPlow’s kind and effective support!"
- Levente Pánczél, Groware Ltd.
"I wish all companies in the computer industry took supporting their products as seriously as DataPlow."
- Bill Arzt