Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[JustinChase]

I've got some weird network issues going on, and in searching for an answer I ran across a thread discussing the value of (and showing how to) setup thumbnail caching on the unRAID server, using MySQL.

someone posted this...

I created a couple of screencasts demonstrating how fast/slow (you be the judge) the thumbnails appear.  For those who are undecided about undertaking this intimidating project.

One when wired:
http://www.youtube.com/watch?v=YItfJxsL0Go

The other when wireless:
http://www.youtube.com/watch?v=9OTb83phqQc

If you click to 33 seconds into the first video, you can see them browse around.  I was floored.  It's quite the experience really.  It's still pretty impressive on wireless in the second video also.

This is not to imply that MySQL is good/bad, just that the experience with the changing art, the views, presentation of information, ticker along the bottom, video/TV continuing to play while browsing was all very good; just a good looking experience I thought.  I think that MC can be/is as fast (I still have my network issues), but the Theater View experience is not as easy/useful in my experience so far.

I'd never seen XMBC in use before.

[perpetual98]

That looks pretty slick.  I never messed around with XBMC for more than a half hour or so.  I just couldn't get it to do what I wanted to do easily.  The last time I messed with it, I couldn't even get it to play my ripped DVD's in folder structure using VIDEO_TS folders.  I think at that point it didn't do recorded TV either.  Not sure if that's changed.

[Daydream]

MySQL helps remembering across all clients what movies you've seen, where did you left a movie if didn't finish watching (stop it in the living room and continue in another room), etc. But it's soundly beaten by a local copy of all art, on an SSD. There is no contest.

A scroll speed comparison MC to XBMC is a bit difficult since each program employs different tricks (MC stores certain 'art' in 3 different resolutions, XBMC may fall back to not showing art to keep the scroll smooth).

Since our focus over here is probably MC (hehe), I have 3 comments:
- max scrolling an album view in Audio (all art stored on SSD) it's so fast it becomes a blur (plus it doesn't look like your navigating, the selection highlight remains frozen in the middle of the screen, while the rest of the imagery becomes a moving haze of all possible coverarts). Maybe something can be further tweak here? As it scroll MC does some fading for top row / bottom row. At supreme speeds this cuts heavily into the visible area.
- the behavior of the existing art for anything not album covers is rather fixed (AFAIK). As an example, the speed the background fades in and out should be much faster or customizable (I can jump in an out of a series even before the background changes; fading should be quick, half a second, not 3-5 seconds like it looks now).
- there is so much more art out there that MC doesn't use yet. As a list from the top of my head as inspiration if any more thought is given to further art implementation, it's like this (if you don't what one looks like visit http://fanart.tv/)


Movie:

Poster
Fanart
Extrathumbs


TV:

Series Poster: Poster, Landscape, Banner
Series All Seasons Poster: Poster, Landscape
Series Fanart
Season Poster: Poster, Landscape
Season Fanart
Clearart
ClearLogo
CharaterArt


Music:

ClearLogo
CDArt
Artist Background
Album Covers

[lhwidget]

I guess I'm lost here.  I run MC16 with the libraries local to each HTPC (3 right now). 

Only the data files are on the server.

I'm using dual core Celerons/Wolfdales and Nvidia 210, 220 or 520 video cards and Xonar DX sound cards in the client HTPCs.

Once I update a library on a machine, the screen response for the thumbnails are about like in the linked the video while they are being generated the first time.  After that, only pulling cached images from faster hardware would be quicker.

The downside is manually backing up the library after importing new material and uploading the new library to the other HTPCs, making sure I have the same version of MC16 everywhere, and making sure I use the exact same drive letter mappings to my various media file shares. 

I haven't been motivated enough to try to automate this, but the responsiveness of the HTPCs is first rate.

I've not experimented with serving the library, is this what you guys are doing?

[Daydream]

I've not experimented with serving the library, is this what you guys are doing?

That too . With my other config (around a Core i5, not SB) and Gigabit wired net, I have no complaints, serving the library and any art it comes with it's very, very fast. It's a couple of orders of magnitude better than my past experience with MySQL/XBMC.

Now, the thing to remember is this is not an apples to apples comparison. Currently I find MC crazy fast, too fast almost in some cases . But look at what it serves, and then compare. It serves album covers and movies/series posters stored locally (very fast), and 'fanart' (backgrounds) retrieved online (with added DL time and a fade-in and out timing that I questioned above). Compare this with the fact that XBMC may display at one single time multiple fanarts for one item which amount to a considerable amount of pictures (some 1080p) that need to be decoded altogether to keep the flow of navigating from breaking up. One 1080p JPG can be decoding really fast on any modern machine. But 10 JPG in a fraction of a second (since I locked the scrolling key and I want the screen to move at max speed). A network setup will suffer. A local SSD will help but there's still some very real I/O limits.

MC is very fast now. Add some of the art I listed above and then I'll be curious how it stacks up.

[JustinChase]

The (big GIANT) downside is manually backing up the library after importing new material and uploading the new library to the other HTPCs, making sure I have the same version of MC16 everywhere, and making sure I use the exact same drive letter mappings to my various media file shares.

Waaayyyy too much work to keep it all straight.

I would prefer to have all the data on a server, and have the clients pull from it as necessary, and anyone updating it will propagate to the other clients, without me spending a few hours updating/copying/pasting/etc.

Again, the video was more of a "look how they do that, it's so informative and vibrant, and useful" and not so much a "look how fast" to me.

[lhwidget]

Point taken.

While what I do takes some planning during set-up, it really isn't so bad, a few seconds to export the library from my main machine where I rip media, and a few seconds to import it the next time I use a client.  Really not painful, just a little bothersome maintenance.

Sorry I missed the point on the vids, I did notice that the performance was quite good for what the system was doing.  It's a good validation for the methodology.

It's just that I have always found putting all aspects of the playback load on the client, and throwing a modest amount of CPU & VPU horsepower at it seems to be the most straight-forward way to good performance for me. 

I'm not familiar with database performance over network optimization and I can barely keep my nose above the software/hardware/OS/networking information waterline at this level in the game 

[JimH]

With my other config (around a Core i5, not SB) and Gigabit wired net, I have no complaints, serving the library and any art it comes with it's very, very fast. It's a couple of orders of magnitude better than my past experience with MySQL/XBMC.

This is one reason that we are unlikely to overhaul the database structure and client/server architecture anytime soon.

[stottle]

This is one reason that we are unlikely to overhaul the database structure and client/server architecture anytime soon.

Would adding additional artwork capabilities (backdrops/series artwork/etc) require an overhaul?  I had been assuming it could be done within the existing system if you guys decide to go that direction (we are hoping you do).  So I'm surprised that this is your reaction, Jim.

[JimH]

Would adding additional artwork capabilities (backdrops/series artwork/etc) require an overhaul?

No.