6 Replies Latest reply on Oct 17, 2011 9:49 AM by Lyle3141

    Media Server - Does it make sense?

    Lyle3141 Level 1

      I'm looking for a DIY solution to allow three work stations to store, access and share media files from a single source.  Currently client project files are assigned to an individual editor and stored on his work station on a RAID0 array.  This is great for the individual editor but creates hassles when projects are re-assigned (requiring moving lots of data) or when someone else needs to access those files in the his absence.

       

      My thought was to build a central repository in the form of a dedicated server to hold all project and resource files and then access and edit projects over the network.  Roughly, my thinking was this:

       

      Server configuration:

      1 x SSD 100GB OS Drive

      3 x 10,000 RPM 150GB Drives for project files (.aep, /.pproj, .etc)

      6 x 7,200 RPM 2TB Drives for media storage in RAID 10 configuration (video, audio and image files)

       

      What I'm wondering is how much of a hit in performance will I be taking by editing over a Gigabit network vs. an Internal SATA III RAID 0 array?  What sort of problems will I likely encounter when three editors are simultaneously putting demands on the server's RAID 10 media storage?  Is this an acceptable solution?  Or have I totally missed the boat?  If I've missed the boat, can someone tell me how larger production houses store and share media?  We're expanding our production capacity and I'd like to have a better solution than three stand-alone edit platforms.

        • 1. Re: Media Server - Does it make sense?
          Harm Millaard Level 7

          Don't even think about a solution along these lines. It is doomed from the beginning.

           

          1. CS5 is not network friendly.

          2. Performance will degrade seriously. Think of a factor 6 or more.

          3. Data integrity is not safeguarded.

          4. Single NIC is way slower than internal raid0.

          5. Reliable switches are expensive.

          6. Good backbones are generally unaffordable for small companies.

           

          The basic problem is that PR is not a multi-user/network capable application. So it is not possible to have multiple editors work on the same source material. You need Autodesk Smoke/Inferno etc. to do that. Starting at $ 80 K.

           

          Consider the following scenario. Two editors working on the same time-line at the same time.

           

          A adds a title to a clip X and saves his file.

          B removes or modifies clip X at the same time, where A inserted the title and wants to save the project as well.

           

          What should be done?

          • 2. Re: Media Server - Does it make sense?
            Lyle3141 Level 1

            Hi Harm,

             

            Was hoping you'd see this.  Thanks for your helpful post on RAID arrays from a couple years ago. 

             

            To clairfy:  Editors would never work on the SAME project simultaneously.  Each of the three 10,000 RPM drives would contain their individual work files and normally each editor would only access their own projects.  The media storage RAID array would contain ALL media files, but again any given media file would only be accessed by a single editor.  Of course three editors might be accessing the media for three different projects at the same time.

             

            Even given this clarification, though, it sounds like I should steer clear of this solution.

             

            So is there a better solution than three stand alone stations that basically live in isolation?  If I re-assign a project do I have to copy several hundred GB of data over the network from one station to another?  If one editor needs to work on a project in the absence of another editor, does he have to work on the absent editor's station?

             

            Thanks again.

            • 3. Re: Media Server - Does it make sense?
              Harm Millaard Level 7

              If you have dual NIC's @ 1 Gb on all connected PC's or use 10 Gb connections, and a switch that supports jumbo frames and you can apply load balancing, you may alleviate the network problem significantly. Just make sure that each editor uses separate projects at the same time. I would keep the media cache and previews locally however.

              • 4. Re: Media Server - Does it make sense?
                Frédéric Segard Level 2

                Lyle,

                 

                As Harm pointed out in his first post, even if you had dual NICs, or quad NICs (better in your case), you'll still need a very good, and properly configured network switch to allow efficient flow of the aggregated NIC bandwidth. Otherwise, it's just money thrown out of the window. But it's doable. Of course, if you could afford 10Gb NICs and switch... even better. Mind you, I'm not sure how the cache database will react when a project created on computer A get's opened on computer B (non simultaneously, of course).

                • 5. Re: Media Server - Does it make sense?
                  JEShort01 Level 4

                  Lyle,

                   

                  Since you should have a good backup strategy anyway, and since re-assigning would most likely not be done in the middle of a work day, why not set up 3 workstations with a automated backup to occur ever day at 3am (file copy style, not compressed traditional backups - aka run Beyond Compare or some other mirroring software as a scripted batch) to copy all of the days work excluding scratch, cache, etc. to the server machine? That way, any of the editors could copy down the files that they may need that someone else was working on. Go with fast arrays on the workstations and the server and with traditional 1GB/sec cabling and you should be able to move things around at about 12GB/min which should not be too debilitating (i.e. 300GB project in 25 minutes).

                   

                  Jim

                  • 6. Re: Media Server - Does it make sense?
                    Lyle3141 Level 1

                    Thank you Frederic and Jim for chiming in.  I like Jim's suggestion as it will more likely fit into my operational budget.  I'll hopefully be moving forward with this within the next 3 to 6 weeks.  I'll re-post to let you know how it goes.