When we think of today's digital studio pros, we don't think network engineers. We envision audio and video artists and editors and other creative types engaged in a common digital content creation task. But there's no collaboration without connection—network connection, that is. Where do today's studio pros find the servers that serve their high-throughput network needs, and how can they build them with a minimum of networking knowledge?
October 2002|The "digital studio" of yesteryear was a marvel of production. Anyone who has watched a documentary on designing aircraft, creating animation, or constructing a skyscraper in the 1930s and '40s would have seen a studio as a vast sea of desks. Hundreds of designers were at work using Stone Age tools creating, editing, and producing final versions of their output. The entire process was a brute-force effort in manual labor of prodigious proportions. It was miraculous that so much came from such simple methods.
Today's digital studio is even more miraculous—bringing creators and consumers within hours or minutes of each other rather than days, months, or even years from concept to content. Thanks to network technology, today's content creators are better able to deliver their visions, at lower cost, and with far fewer people.
Key to this efficiency and effectiveness is the inclusion of network servers into the workflow. Servers and the digital studio are sort of axiomatic. You can't do much now in audio or video production without a server, and that trend isn't likely to change, with continuing growth in the number and types of servers that most networks will have in the near to middle term.
To Connect and Serve
It isn't a question of whether you have a server in the system—you already do. Even single-person shops also have "servers" in the form of Windows 2000 or XP. The process of "logging in" as a specific user on a Windows XP PC is in effect "logging in" to a server with this single workstation effectively an "integrated" network server.
The value of servers becomes clear with a bit of discussion. Servers by definition provide two kinds of services to the digital studio:
1. They can control and provide access to content.
2. They can generate new content.
While either service could be done on any PC, the network server "democratizes" the service so all can enjoy its benefits. This was an incredible leap for users in 1985 when network servers first became a buzzword. It still gives incredible value to designers today.
Novell NetWare and Windows NT are most widely known as "file servers"—providing access to content. This access could be as digital information over the network (local, intra-company, or beyond). It could also be as hardcopy "access" or on-screen display.
Database servers, message servers, and encoding servers go a step further as they create new content (and likely provide access to that new content as well.) These can be MPEG streams, MP3 audio, or HTML output.
The good news is that the industry does recognize a difference between the typical office automation system and that of the digital studio. IDC, one of the largest of the analyst groups for the IT industry, identifies the tools of the studio as "The Rich Media Asset Management Market." This market focuses on the technologies involved in creating, acquiring, managing, filing, securing, and delivering digital assets. Digital content includes audio, video, text, and all kinds of non-static imagery, regardless of the medium.
(Curiously, IDC does not consider still images as part of "rich media" although many studios spend considerable time on such images for box or cover art, presentation slides, or billboards and displays.)
The critical issues facing the administrator today in the digital studio are:
• "How do I handle growth?"
• "How do I increase the number of services to our users?"
• "How do I decrease costs, improve reliability, and increase performance—all at the same time?"
The choices and constraints are far from obvious and become explosively complex as the network expands. Special purpose or dedicated servers do provide the quickest way to handle growth, increase services, and improve reliability by assigning specific tasks to such boxes. Unfortunately, defining these tasks creates an additional set of issues.
Envisioning the Video Server
The type of special-purpose server that figures most prominently in the digital studio is the video server. The term "video server" should clearly be found in any digital studio's lexicon, but industry pundits can't quite decide what exactly a video server is. To serve the digital studio effectively, the video server should provide one or more of the following five features:
1. Real-time MPEG-2 encoding of content.
2. Real-time MPEG-1 encoding for surveillance and monitoring.
3. Content management services for an attached RAID or optical library of digital content for local access.
4. Production support for storing and retrieving work-in-progress content for local access.
5. Streaming video to the Internet.
Each of these tasks requires different hardware, operating system optimizations, third-party software, and management tools to cope effectively with growth and easy access to the new service. In effect, these are different systems all under the same moniker of "video server."
Some vendor offerings do answer these complexities. IBM's impressive Digital Content Creation Initiative—including their Media Production Suite for Video Producers and Custom Digital Content Management system—while supporting these services, appear to be tailored for an Enterprise or Fortune 100 or large studio deployment. Fortunately, though, for the small- or medium-size studio, vendors are addressing these complexities with lower-cost NAS, SAN, and new management protocols to try to approach this in a more "plug-and-play" fashion.
In fact, one of the hurdles is in realizing that, for network services, "more is less." More hardware and more devices on the network can be much less hassle. In the past, the high cost of the network's primary server (due to the need for a higher-performing system) meant that administrators wanted to get the most out of one piece of hardware. With the types of systems available today, this approach misses the mark by a mile.
Today's 1gHz systems cost well under $500 (without monitor), so there's little reason to try and force a server to take on more tasks. One of the reasons that Windows NT gained a reputation for unreliability was its users' tendency to overburden NT servers. If an NT server only had to be an Exchange server, then it could be optimized for that task and expected to perform with a minimum of reboots. Unfortunately, if that NT server also had to provide end-user file access, print services, directory services, and the like, then it stood a pretty good chance of succumbing to schizophrenia.
Also keep in mind when assigning server tasks that video servers aren't the only type of server in demand at a digital studio. An office automation server—the more typical Windows Exchange or Microsoft Office file server—provides a medium-duty service with less demand on throughput.
Today's digital studio pros can choose among both software and hardware packages that will create a video server providing any or all five of the preceding services. These solutions vary substantially in price and performance. Who you are and what you need dictates the type of solution you'll choose. Thus, here another aspect to server selection comes into play—who is the audience?
Digital studios encompass perhaps four general classes of users: designer, accounting, external partner, and consumer. What type of server best suits each task? Designers, who need to store production multimedia files, need internal network access via Ethernet and/or SAN using a production server. In accounting, we need to track project status and invoice totals; our ideal server set-up is an internal network using an office automation server. External partners need to track project status, which means they need Web access to an office automation server over a Virtual Private Network (VPN); external parnets also need to view completed multimedia content, edit existing content, or generate new content, which means they need Web access to a streaming server over a VPN. Finally, consumers need to view completed multimedia content, which means they need Web access to that content via a streaming server, or the ability to download content from a Web server with the option of purchasing it in hardcopy.
An administrator should survey the need for each class of user to determine the impact on server design. Say you've got a 40-person animation team working on deadline for a release date. Some are doing real-time IP transmissions of MPEG content, others are transmitting digital rushes from one location back to the studio, and they all require significant server resources. However, the former require significant local access (big RAID on SAN) while the latter require a good pipe (a T3 connection or better).
Administrators need to survey the content types generated or accessed for video server deployment. Consider the disparity in average file sizes, ranging from 80KB (Web GIF) to 3D animation (60MB/frame), with MPEG-2 (3MB/minute) and MP3 audio (at 1MB/minute) somewhere in the middle.
In fact, storing just one second of MAYA animation could take 1.5GB if each frame were different! This type of video file would have been a major challenge just a few years ago when the NetWare file system (for example) could not handle files larger than 2GB. (With the introduction of the Novell Storage Services—NSS—in NetWare 5, this limitation went away.)
More NAS for the Masses
One way vendors solve the challenge of server deployment is through pre-configured systems like Network Attached Servers (or Storage—NAS). NAS uses the network cable (likely a Cat5 or Cat6 copper or fiber) as the high-speed bus to connect servers to users. In the future, the iSCSI protocol will unite various NAS servers into one virtual unit.
Already, conventional Ethernet NAS units can work together as one virtual volume (for example, Maxtor's). One highly effective NAS solution has appeared in Apple's new Xserve rackmount server. By creating virtual file systems for Unix and Windows users, Apple can co-opt these to use a Mac-based system. Since management is done through Web-based tools, end-users and administrators need hardly know that there's a Mac on the network. Throughput on the Xserve is expected to meet digital studio demands, even in its initial two-CPU configuration.
Other NAS-based services are also beginning to appear. DinoStor, Inc. recently introduced their DinoStor Tape Server. This NAS unit creates a backup pool from existing SCSI-enabled tape libraries. The unit has both 10/100 and Gigabit Ethernet onboard as well as U160 SCSI connections.
The glue for the DinoStor NAS server is the new Network Data Management Protocol (NDMP). Servers and workstations that support NDMP can readily communicate with the DinoStor server as if its backup services were local to that workstation or server. This high level of automation is critical to any studio. Currently, over 60% of mission-critical data sits unarchived on local workstations, laptops, or even servers. Requiring user intervention to initiate, monitor, and terminate backup sessions has been a 20-year effort in futility, as any administrator knows all too well. If NDMP rapidly proliferates, the DinoStor solution should go a long way towards fixing that problem.
The DinoStor system is only the first of several potential Network Attached Services to come. Keystone Learning, a Provo, Utah-based training company, typically provides its courseware via traditional CD or DVD discs. However, it has a new "Learning Library" NAS with preinstalled courses already on the device's hard drives. Keystone's innovative Learning Library means a five-minute install rather than have to perform a network install for each class, or maintain a physical library of discs for sharing, or to install an optical jukebox to share the CD- or DVD-ROMs.
Rackmount NAS devices, like all PC hardware, will continue to plummet in price while increasing in speed and capacity. (Currently, 25GB hard drives are more expensive than 60GB drives.) Digital studios may well opt for libraries of stock photos, footage, models, and the like, delivered via a pre-installed NAS unit to plug into the network cable if this only adds 4-5% to the overall cost of the library.
Ituner Networks has developed one type of NAS video server—the MediaBox VS-2601 Streaming Video Server. It incorporates four Shoutcast servers, IceCast, QuickTime (QTSS) and Real (with the appropriate license) on a Linux chassis. Set-up time is only a few minutes for the server itself, although the Internet connection and video/audio feeds need to be there already. The MediaBox eliminates a significant amount of design and experimentation for the administrator for only $4500, a relative steal in comparison with consulting time.
Digital Asset Management, Studio-Style
We mentioned earlier IBM's comprehensive Digital Content initiative. Steve Canepa, IBM's vice president, global media and entertainment industry, is at the forefront of meeting the needs of digital studios. As he outlines it, the three key components of content development—creation, management, and distribution—are evolving from separate compartments with manual (meaning humans carrying things around) operations to an automated, much more efficient single operation built around a common set of com- munications services. The artist, sales force, and end-user will now have real-time interaction on each project, he says. This significantly reduces the amount of time it takes to conduct business together. It also reduces the error rate since fewer people have to act as "middle men."
"In the past, vendors tried to answer this need with just hardware or software," says Canepa. "For the digital studio, that simply wasn't enough. It takes three things: hardware/software, consulting services to tailor the solution to specific needs, and new technology to better answer those needs."
Canepa's answer to the technology component is IBM's new Global Parallel File System (GPFS). Traditional file systems, like Microsoft's NTFS, are built around the needs of office automation and database operations. This means that such file systems work extremely well for small (i.e., less than 50-100KB) file sizes. Unfortunately, for multimedia and video files this introduces significant overhead and latency into the equation. Rather than try to patch a digital studio "enhancement" into this kind of file system, IBM created a new system which supports digital content needs. GPFS thus provides:
• Up to 500TB file sizes
• Support for 12GB/sec transfer rates
• Full duplex (simultaneous read/write) operation IBM also has honed a specific line of workstations for the digital studio, the Linux-based Intellistation.
Canepa feels that this suite of products and technologies creates a solution that is better than the sum of the parts. He describes one studio using this suite that no longer has an in-house design team. Instead, outside contractors generate all the digital content remotely via IBM's content management services and then transfer the completed materials electronically to the client. By "virtualizing" the studio, costs are reduced and more projects can be taken on. Canepa sees this as a major step forward for the digital studio and its ability to compete and complete projects.
Building the Digital Studio Network
The crux of today's networking is this: many more creators can now spend time creating and less time administering. Users also spend less time searching and more time enjoying content, thanks to better-performing networks.
SIDEBAR: An Open-Source Future?
It used to be that high performance meant a Unix-based server, using an RISC processor, with data stored on a SCSI drive. Today, the performance of Intel or AMD-based Windows servers using EIDE/ATAPI drives approximates that of high-end Unix workstations at much lower cost.
An additional movement has begun to move development to Linux-based servers and workstations. This is happening for various reasons:
1. As a cost-cutting measure, because Windows licensing is increasingly expensive.
2. As an alternative to rip-out-and-replace, seen most recently with the introduction of Windows 2000 and Active Directory. (To gain any real benefit from the directory, every server and workstation had to be Win2K. This likely meant upgrading hardware and software—i.e, replacing everything except user data.)
3. As a way of "future-proofing" access. Given that Microsoft Office file formats have radically changed over the past several versions, it is possible that future versions may not even be able to open older saved Office files. This raises the concern about how to archive all existing content for future access.
However, Linux—either as a more common server operating system or as a desktop replacement—is still a year or more from widespread adoption. Linux equivalents need to be truly "equivalent" in file format and usability in order for users to migrate. Even if file formats are faithful across multiple platforms, the user interface for many Linux tools and software are primitive in comparison to the ubiquitous and familiar Windows conventions.
Companies Mentioned in this Article
Apple Computer, Inc. www.apple.com
DinoStor, Inc. www.dinostor.com
IBM Corporation www.ibm.com
Ituner Networks Corporation www.ituner.com