IT Trends: Opening the Data Dam
Mar 1, 2004 12:00 PM,
By Brent Harshbarger
Media requires a great deal of distribution bandwidth, but the story doesn’t stop there. Moving data around is one thing, but you also need a place to store it and a way to retrieve it. Ten gigabit Ethernet will soon make asynchronous transfer mode an endangered species for wide area networks (WANs) and even metropolitan area networks. Once this technology takes over, the “data dam” will burst wide open, and it will become much cheaper and much easier to implement to a point. However, it’s not enough to have a large distribution pipe and then have bottlenecks at the storage point. That is like putting a garden hose and a bucket at the end of the Alaskan pipeline. Because of this issue, specialized storage and retrieval technologies for video servers have been developed and deployed. These include SCSI and Fibre Channel.
Fibre Channel is a networking technology that was developed to address the speed, capacity, and reliability requirements associated with communication between storage and server devices. Fibre Channel development began in 1988 and became a standard in 1994. Since then it has been widely used in media systems because of its wide bandwidth and its ability to move large amounts of data. The disadvantage to this technology is that it requires specialized interfaces and switches that increase the cost of a system. Fibre Channel also requires more attention to design details if it is required to communicate with the local area network and WAN technologies.
Even with those limitations, Fibre Channel became the driving force behind storage area networks (SANs). It is widely deployed as a 1 Gbps (with 2 Gbps soon available), but with gigabit Ethernet and now 10 GbE, that may be a moot point. The raw bandwidth of the 10 GbE, which can be easily scaled, offers a cost-effective advantage that may soon put Fibre Channel in the “remember when” category. It would have already done that if Ethernet had provided a better means of dealing with large blocks of data and reliability issues. However, in recent years Ethernet has been getting its act together. Not only has it been boasting higher levels of reliability, but coupled with IP technology, Ethernet can provide a robust interoperable and scalable framework for large media systems.
Moving Fibre Channel from 1 GB to 2 GB is a move to try to stay competitive. However, with such widespread deployment of Ethernet and its 1 GB to 10 GB, it will take an astronomical leap for Fibre Channel to even maintain its current status.
ISCSI TO THE RESCUE
Moving data efficiently in and out of storage systems requires a technology that will connect the physical storage system with the distribution network. For many years now, small computer systems interface (SCSI) protocol has been the de facto standard for accomplishing this task. Its claims to fame are its high data transfer rates, reliability, and low latency. However, it has not worked well with Ethernet, which is another reason Fibre Channel has remained the standard storage system, until now. A new version of SCSI has emerged that will allow Ethernet to replace Fibre Channel in storage systems for media and other high-volume data systems. It’s called Internet small computer systems interface (iSCSI), and it’s an Internet Engineering Task Force standard that maps SCSI blocks into Ethernet packets. Basically, the iSCSI protocol is a method for transporting low-latency SCSI blocks across IP networks. The advantage of iSCSI is that it lets you build SANs over IP using something you already know and use every day: Ethernet. In this way, iSCSI removes the limitations of direct attached storage (DAS), including the inability to share storage resources across servers. But to fully appreciate iSCSI, some background information on storage systems is required.
Storage technologies started out like most other computer technology: as all-in-one centralized frameworks. Over the years, it has moved with its relatives to become more decentralized and then distributed. There are three main storage technologies widely used today: DAS, networked attached storage (NAS), and the youngest of the family, SAN.
DAS is simply where each server has dedicated storage. Storage is seen and accessed by one host system. Should another host system need more storage, it will have to be physically added, perhaps along with I/O interfaces or host bus adapters. When thinking about a DAS system, the first thing that comes to mind is an old mainframe system that used reel-to-reel tape. However, a PC that has a hard-disk drive locally can also be considered to be a DAS. Therefore, it is still a valid technology today, but larger and networked systems require remote access.
NAS is shared storage that is attached to a network. NAS enables data and storage sharing by adding a layer of intelligence on top of a host system or special server to provide and control access to the storage media through standard network interfaces. Host systems access this shared storage via standard software. NAS has been a popular model for traditional file servers during the past decade or so.
SAN is a network of shared storage devices. A storage area network describes a dedicated, high-performance network infrastructure deployed between servers and storage resources. The storage area infrastructure is a separate, dedicated network entity optimized for the efficient movement of a large amount of raw block data. In effect, SAN is an extended link between server and storage and enables the extension of the SCSI protocol over longer distances. SANs enable storage to be accessed using networks rather than dedicated interfaces like parallel SCSI. They have the ability to support multiple mixed operating systems, use open standard interfaces such as Fibre Channel, support multiple types of media and topologies, and support multiple protocols including SCSI and TCP.
The combination of the new higher bandwidth Ethernet standard along with iSCSI integrated with SANs provides the digital media professional a powerful arsenal of tools. As digital content grows, it will require larger and faster cataloging and retrieval systems to support it. For rich media technologies and services like video on demand, HDTV, or digital cinema to be successful, a storage system matched with proper network bandwidth is required. Ethernet and iSCSI provide a cost-effective and efficient infrastructure for the systems integrator to deploy production and distribution systems. These technologies provide the path to the promise of convergence of the intranet and interactive rich media to provide real-time content and communication.
Brent Harshbargerhas worked for Peavey in the development of MediaMatrix. He can be reached at[email protected].