Storage technology
By Greg L. Doyle
NAB2002 provided an incredibly diverse line of storage solutions for broadcast, production and enterprise data storage. The gamut ran from workstation-attached FireWire (IEEE 1394) drives all the way to NAS and SAN solutions for the enterprise. Storage is a complicated issue with tradeoffs ranging from performance to reliability to scalability to manageability and most importantly, price. These issues are spread across hardware, software, service and support. Today's high-performance storage market offers three basic options: direct-attached storage (DAS), network-attached storage (NAS) and storage area networks (SAN). Other interfaces, such as FireWire, provide a serial interface to either standard IDE or newer ATA serial interface to standard PC drives but are essentially DAS.
Workstation-attached storage
Direct-attached storage has been the most common method of attaching high-performance mass storage to workstations, servers and local area networks (LANs). In its simplest form, DAS consists of a disk drive attached directly to a computer through SCSI commands, the most common means of I/O communication between a computer and a hard drive. SCSI transfers data as blocks, which are the low-level, granular units on storage devices.
High-performance storage drives incorporating SCSI connectivity saw a boost in performance this year with implementation of the seventh generation of SCSI I/O: Ultra320 SCSI. The maximum sustained transfer rate of SCSI drives has been increasing to meet the need of modern systems and applications for higher data throughput. Because SCSI allows data to be striped across multiple drives for an increased aggregate bandwidth, a rule of thumb has developed that the SCSI bus bandwidth should be at least four times that of a hard disk drive's maximum sustained data rate. A year ago, a typical SCSI hard drive could theoretically sustain about 55 MBytes/s. New generation drives can sustain close to 70 MBytes/s, therefore, as few as three drives could saturate an Ultra160 SCSI bus. Ultra320 SCSI is projected to provide the interface data rate necessary to satisfy this bandwidth requirement for at least another year, and it's still backward compatible with earlier versions of UltraSCSI interface.
A RAID-5 array with Seagate Ultra320 disk drives connected to a two-channel Adaptec Ultra320 SCSI RAID controller will run multiple streaming video applications with data flowing at 320 MBytes/s across the SCSI bus. Also, in the Ultra320 SCSI controllers, LSI Logic's LSI53C1030 PCI-X to dual channel Ultra320 SCSI controller was featured showing LSI Logic's integrated mirroring at Ultra320 SCSI transfer rates to Seagate's Cheetah X15-36LP enterprise drives. LSI Logic's Fusion-MPT based controllers provide interoperability. The Seagate X15 drive operates at 15 K rpm and delivers seek times as low as 3.6 ms and latencies as low as 2.0 ms.
Storage area networks
SANs are dedicated networks that connect servers to storage devices and transport storage traffic without burdening the enterprise LAN. SANs can be attractive for several reasons, including high performance, reliability, availability and scalability. SANs also are useful when you require large amounts of data backup because high-density storage solutions, such as tape backup or DVD jukes, mount directly to the SAN.
Unlike network-attached storage, a SAN will provide centralized file management and limit redundant copies or multiple versions of files, which can rapidly consume disk space. If a system is not in place to manage this, servers with high-demand applications and often-used data can become overloaded, while other servers sit idle. SANs help eliminate this problem and improve the efficiency of the LAN by moving the constant stream of storage traffic off the LAN over a host bus adapter (HBA).
The HBA is connected via switched Fibre Channel to the storage device, and a separate file management system tracks metadata related to the files and updates clients on file status. This metadata establishes file access priorities to limit who can write to a file and who can read a file. It would be disastrous to have two people attempt to simultaneously write to the same file.
SANs are built on high-performance RAID arrays that are raw-block devices. These blocks can be very large depending on the storage platform. Users carve up the terabytes of data into hundreds of gigabyte-size chunks called Logical Unit Numbers (LUN) and assign them for private and exclusive use by the individual users. Each server, in turn, places a file system on the LUN, but the server's storage is block-level.
Connectivity between the storage devices and the workstations is achieved using Fibre Channel, either as an arbitrated loop or as a point-to-point switched fabric. This year, SANs saw a boost in performance using the Ultra320 SCSI drives and 2 Gbit Fibre Channel. This combination was shown at several high-end storage vendors, including EMC, LSI Logic, Rorke Data, SGI and StorageTek.
This year, SGI celebrated its 20th year in business and its 10th year as an NAB exhibitor. The company showed the CFXS2 media server, allowing a multi-OS infrastructure to enable sharing of files between different operating systems. Currently supported operating systems are IRIX, Solaris and Windows NT.
StorageTek introduced the D178 disk subsystem through a strategic alliance with LSI Logic. The enterprise disk subsystem supports more than 40 TBytes of information. StorageTek's 9840 tape drives are now delivering 20 GBytes (native) per cartridge allowing a full petabyte (1000 TBytes) in a tape library. Also, the speed to data is 12 seconds, and acts as an alternative to an all-disk implementation, which can be costly. The 9840 can deliver throughput of 19 MBytes/s native and up to 37 MBytes/s compressed. The SN6000 with Virtual Transport Manager (VTM) provides a single storage image for a data center's open systems environment. Unlike conventional server-based and storage-centric approaches, this network-based approach frees application servers from storage management overhead.
DataDirect debuted the S2A 3000 Silicon Storage Appliance, a 1RU SAN appliance that allows users to plug the workstation into one port and a commodity priced JBOD into another. The unit can be configured with up to 7 TBytes of storage.
Avid is now providing a direct I/O device for the Unity line of servers with the Xdeck. The unit incorporates SDI I/O, FireWire support, RS-422 remote control and VTR-like control functions for ingest and playout directly from the Unity server.
SeaChange unveiled its new Broadcast Media Library 2400 server. SeaChange's BML provides online availability exceeding 99 percent. The architecture is different than a traditional SAN with the Fibre Channel fabric spread across multiple nodes in a three-dimensional, toroidal pattern, thus eliminating a single switching point.
SeaChange discussed releasing a 2.5 Gbit bussed architecture for storage access soon. This will be a proprietary platform that probably will be something similar to Infiniband, where the movement of blocks of data is off-loaded from the CPU bus to peripheral hardware devices.
LSI Logic had several high-end storage solutions on display at the show. The MetaStor E4600 storage system features 2 Gbit connectivity, 40 TBytes of storage and up to 390 MBytes/s of sustained throughput from disk to sustain multiple video streams. The E4600 HPCx combines multiple E4600 systems and one or more SAN expansion modules (fabric switch) into an integrated solution. All MetaStor systems offer a modular design and robust SANtricityStorage Manager software. SANtricity-powered MetaStor storage systems support mixed RAID levels, drive capacities and rotational speeds, and individualized volume settings to ensure attached host systems receive their desired storage requirements. And SANtricity provides automated path failover and online configurations, reconfigurations, expansion, maintenance, and performance tuning to ensure valuable data is always available.
Network-attached storage
Where a SAN storage model provides direct access to raw blocks of data, a NAS provides a file system that manages the blocks of data in a file format and provides an embedded engine close to the disks, which represents the disks as a “network file system,” or NFS. Hosts see it as a network file system, not a block device. The network-attached model offers ease of administration. It can mean fewer file systems for an enterprise because many servers can simultaneously share each file system. In a network-attached model, an administrator could have a single file system for a terabyte-size NAS device. The entire terabyte would be available to every server at any time.
There are limitations in the areas of performance, impact on host processing, scalability and “unsettled writes” that prevent NAS from being effective in some applications such as multimedia. Host impact refers to how much overhead the LAN traffic erodes in application host CPUs. Moving multiple megabyte-sized files to or from a LAN is much more expensive than moving data from a direct-attached pipe (SCSI, Fibre Channel). The scalability limitations are due to CPU performance and the physical makeup of the LAN. Across a LAN, traffic must be processed at both the application host and the NAS server end. “Unsettled writes” can be caused by the fact that LAN-based writes are not 100 percent ensured to be settled out to real storage at the moment they are acknowledged. Once data is moved out to the LAN, the application is typically told it is free to continue its job, however, when capturing live video as a constant stream you need 100 percent assurance that the data is recorded or you'll start dropping video frames.
NAS servers are seeing great improvement in performance by applying multiple CPUs or parallel processing along with gigabit Ethernet and 10 Gbit Ethernet for faster throughput across enterprise or wide area network (WAN) connectivity.
Ciprico demonstrated its new DiMeda 2400 server incorporating gigabit Ethernet to the workstation in a news editing demonstration. Ciprico states performance levels exceeding 120 MBytes/s aggregate bandwidth while concurrently serving multiple application clients running Windows, MAC or Unix operating systems. Because the server utilizes file access technologies, it is application independent.
Avid introduced the Avid Unity LANshare, an entry-level gigabit Ethernet system with up to six dual-stream ports or 10 single-stream ports with up to 640 GBytes of storage capacity.
SAN-attached file system
A SAN-attached file system has the potential to bring the best of both worlds to the workstation. These systems can be expensive, but they allow the application to request data at a file level and receive it at a block level. In SAFS architecture, one server performs as a NAS with a LAN connection to the application server or workstation. The back-end of the NAS server is attached to the storage elements, or the SAN, using Fibre Channel or SCSI. Another path to the SAN is provided to the application server via Fibre Channel to an HBA on the server's PCI bus for block-level access. Applications on the hosts believe they are talking to the file server, which they basically are. The file server performs authentification and layout of the file on the storage, as well as all other NAS-type functions.
EMC displayed such a system using the Celera NAS server in conjunction with the Symetrix SAN and an application called Celera High Road. As large files are requested via the Celera server, the system automatically opens up the second path directly to the SAN for block-level access.
Internet SCSI
A few vendors were talking about incorporating Internet SCSI (iSCSI) for WAN connectivity to storage area network platforms. This will become important as workgroups are spread across multiple facilities such as a centralcasting model. iSCSI is a software package that emulates SCSI protocols, but the connection method is via an IP network instead of a direct SCSI-compatible cable.
The promise of iSCSI is that storage management software, which was originally written for the well-established SCSI standard, can now be used to make a remote disk or tape drive on a network operate just like a local disk. Block-level access can be achieved across a LAN such as Ethernet or even the Internet. The potential benefit is that users can connect to remote storage devices to replicate data without having to write huge amounts of new software.
FireWire frenzy
The popularity of FireWire-attached storage has been incredible. While initially aimed at the prosumer market, this storage technology has found a home in audio and video editing platforms and content capture. With a theoretical bandwidth of 400 MBytes/s, this platform can provide an inexpensive and portable storage solution.
At NAB2002, a few FireWire storage platforms were based on stand-alone operations, such as the Data Video DV Bank FireWire recorder. This system operates on a self-contained processor eliminating the need to tie the storage to a laptop: You can connect the camera or video mixer directly to the DV Bank and start recording. The device also has an RS-422 interface for connectivity to an NLE suite and VTR-like controls on the front of the box. The unit comes configured in a 60 GByte or a 100 GByte drive.
Glyph Technologies exhibited several FireWire storage options for audio workstations and small- to medium-range NLE workstations. The companion storage platform for audio workstations is optimized to reduce noise and vibration to help keep it quiet in the audio suite. FWHS is a 1RU, two-drive FireWire storage unit with up to 240 GBytes of hot-swappable storage. The DV Project is a FireWire RAID drive with up to 240 GBytes of storage at 40 MBytes/s of throughput. This would be great for those using medium-level editing workstations from Apple's Final Cut Pro, Avid DV Express or Media100. If your projects get larger in scope, you can stack more of the DV Project drives and stripe across them for up to 1 TByte of storage.
While some lower cost storage systems provide reasonable to high performance, you should also do some risk assessment to determine which platform will best suit your needs. Just because a drive performs to your bandwidth needs doesn't necessarily make it the right choice. When shopping for a storage solution, consider what the application will be and what the business model can justify. Also consider your legacy systems and determine if newer proprietary systems will play well together with what you already have. As storage technologies evolve, it looks as though NAS and SAN developers will come up with ways to make their products look more like each other. When you implement your storage platform, you need to determine your storage and filing applications for the long run.
Greg L. Doyle is president of Doyle Technology Consultants.
Do you have a comment about this article? To tell us your thoughts, click here.
Back to the top
Next page
Return to Broadcast Engineering
|