Is LinuxMCE for me?

[wierdbeard65]

VLANs are not an option, unless I replace all my switches, as they do not have that capability. Also, in m experience (and one of the larger government agencies of Sweden, which I will not name here), the more configuration options you have on a switch, the more problems you have. Nothing to configure, nothing to go wrong.

Interesting. My experience (at a number of sites, large and small) is the opposite. No configuration = no control = problems. For example, not being able to turn off STP can (and for me often does) lead to huge problems with DHCP. Also, if you are intending to stream real-time stuff over a large (and presumably fairly busy) network, QoS becomes important, particularly if you have a mix of network speeds. Cheap switches often don't support QoS. I had assumed that, given what you said about your network (reliability etc), you would be using commercial-grade managed switches, not the unmanaged SoHo variery.

Be warned, though. Thom, in particular (and to be fair he's right and others will say the same) will tell you not to fight the way it's designed....

Well, if it has to be a choice between fighting the way LinuxMCE is designed and the way my entire network is designed, I'll take the fight to LinuxMCE (or simply use some other media option). I need my network design more than I need a specific media center. No offence intended, it's just the way it works for me.

Fair enough and, as I said before

I understand you may well have a solution worked out that meets your needs and if so, great!

however these forums are searched by many users (both new and old) and a complete discussion is, IMHO, always warranted for the benefit of those who come after 

[trentend]

VLANs are not an option, unless I replace all my switches, as they do not have that capability. Also, in m experience (and one of the larger government agencies of Sweden, which I will not name here), the more configuration options you have on a switch, the more problems you have. Nothing to configure, nothing to go wrong.....

Woah, tiger!

That's just not correct.  For a start managed switches would allow you to trunk up your current small number of available cables to provide a backbone for good bandwidth around your physical network.  Secondly it would allow you to determing what traffic gets prioritised and routed where.  Managed switches would be really useful in your situation.  Some layer 2 gigabit managed switches with all the functionality you might need are quite low cost compared to the setup you have. It's a no-brainer.

The key component in my network infrastructure and separation of subnets is my d-link gigabit managed switch.  I also use them for work.  Not had a single problem in two years.

[skeptic]

I'm not quite following you as to why you cannot separate the DMZ type servers from regular old desktops/media directors.  Are you saying you have servers scattered around the house, sometimes sharing a network drop with what would be a media director?

If this is the case, it's still easily doable without a spof (for the servers).  Just add the lmce core in parallel.  Use static IPs for the servers/printers and lmce based dhcp for the media directors/desktops.  Your desktop type devices will go through the LMCE core (spof), but the servers will still use the existing paths out.  It's not a true switched VLAN, but if you use a completely different IP range for the servers or other static IP non-LMCE type devices then it will function like one (other than not physically separating network traffic).  Even if you just hardcode different default gateways you can avoid LMCE as the spof.

Oh, one other thing I didn't see mentioned, leave the default IP range for lmce as 192.168.80.x.  If you try messing around and giving it a different ip range you will be causing yourself headaches.

[Troberg]

Also, if you are intending to stream real-time stuff over a large (and presumably fairly busy) network, QoS becomes important, particularly if you have a mix of network speeds.

It's all gigabit now, except a few game consoles. I'm alone on the network, and the busy server to server traffic is on a separate network just for them (they are all in the same room, so that's easy). QoS is not a problem. I choke my disks before I choke the network.

I had assumed that, given what you said about your network (reliability etc), you would be using commercial-grade managed switches, not the unmanaged SoHo variery.

I did use them, but honestly, I found them more trouble than it was worth, even though I hardly had any use for their features, so when I switched to gigabit, I went for simple unmanaged switches instead, putting all the logic into the computers, turning the network into "stupid cables" that don't do anything "smart". A nice side effect was that it was also much cheaper.

however these forums are searched by many users (both new and old) and a complete discussion is, IMHO, always warranted for the benefit of those who come after

Of course.

For a start managed switches would allow you to trunk up your current small number of available cables to provide a backbone for good bandwidth around your physical network.  Secondly it would allow you to determing what traffic gets prioritised and routed where.

As stated above, I have no performance issues on my internal network, just my internet connections, and I'm pretty much screwed there until they run an optic fibre or two by my house. I'm nowhere near choking the network, even if I have several servers moving files over it. This is partly due to it all being gigabit, partly because I've used slower 5400 rpm disks for cooling reasons (that's really a problem, in the summer, my server room temp tends to rocket) and because I find them more reliable. I need lots of reliable storage, not speed, and the hardware I have delivers that.

The key component in my network infrastructure and separation of subnets is my d-link gigabit managed switch.  I also use them for work.  Not had a single problem in two years.

D-Link is not a favourite of mine, they tend to overheat and start sending garbage, so I've switched to cheap NetGear at home (never failed yet, and even if they do, I get two for the price of one brand switch) and HP at work.

Once again, our needs are different. Regardless of how it may seem, my network environment is far from haphazard, it has grown and adapted from a need, including two total rebuilds during the last ten years (I've taken the chance to do it when I've moved).

I'm not quite following you as to why you cannot separate the DMZ type servers from regular old desktops/media directors.

It would be a single point of failure, and it would make it difficult to load balance my internet connections. It would also require a major rethink in the network, which is always a potential risk.

I also have a feeling that it might mess up remote administration of the servers, and I do not want 15 monitors and keyboards in the server room...

If this is the case, it's still easily doable without a spof (for the servers).

True, but there would be two problems:

* If the spof goes down, I will not be able to reach the servers, so it will not help me much that they keep running.
* I don't like the idea of piping all traffic between file servers and clients through a single machine.

Oh, one other thing I didn't see mentioned, leave the default IP range for lmce as 192.168.80.x.  If you try messing around and giving it a different ip range you will be causing yourself headaches.

Well, that's some headaches that I'll need to get anyway in that case, given that my network is currently configured for 192.168.0.*, including a bunch of servers which are not on DHCP (Smoothwalls, development servers, DB servers, the inside IF on the web servers and mail server and probably some other servers as well). I like having servers on a fixed IP, as it gives me a chance to reach them, even if name services should die. Reconfiguring those will be a pain in the ass anyway.

My plan is to use a class B address range, using, for instance, 192.168.1.* for DHCP, and 192.168.0.* for static IP's, but letting them all coexist on the same network. It's a little bit cleaner than today, where I, for historical reasons, use 0-16 and 240-255 for static adresses, and 17-239 for DHCP.

[wierdbeard65]

Now, if it only could work as a virtual CD server as well...  I'm a programmer, and if there is one thing I've learned, it's that there is always one more idea to add. The trick is knowing when to stop.

Ahhh, you nearly got away with sneaking that comment in, but I caught you out!!!

Welcome to the development team 

[skeptic]

Troberg - I don't think you fully read/understood my last post.  I/we also don't fully understand your setup.  I.e. which servers are internal vs. external, what spof path are you concerned with, etc.  When I said servers, I was referring to those servers that are accessed from the Internet, webservers and such.  In your reply you appear to be concerned with desktops reaching servers.  Clearly we are referring to two different server groups.

I am still confident that there is a fairly straight forward way to add LMCE into your LAN without major issues.  However, I think we (I) need a diagram to better understand exactly how things are interconnected, what services the servers provide (internal fileservers vs externally accessed webservers or whatever), etc. 

[Troberg]

Welcome to the development team

Sorry, but at the moment, I have my own programs to add insane amounts of useless features to. 

I/we also don't fully understand your setup.  I.e. which servers are internal vs. external, what spof path are you concerned with,

There is no clear distinction between internal and external servers. The servers which are accessible from the outside also have an internal NIC, which makes them a lot easier to manage. As I have them pretty much locked down and behind strict firewalls, I don't see this as a major security breach. In other words, they have two connections, one to the firewalls, one to the LAN.

Your suggestion would create a bottleneck between servers and desktops which would be a spof.

What I want/need is a way to have a nice integrated media solution without having to rebuild/rethink my network, at least not in a major way. I don't get why it have to be so hard, is there really any reason LinuxMCE actually needs to control your network and determine the design of it? The only reason I can see is PXE, but any DHCP server should be able to handle that. I do modern multi tier software for a living, and usually it's just a matter of configuring a few addresses so that the parts find each other.

The more I look it, the more I think that some other solution might fit my needs better, such as XBMC. Sure, it will take a bit of work to integrate my alarm sensors, but that's a later project anyway. At least that's just a simple install, and no need to rebuild/rethink my network. Too bad, LinuxMCE really has some nice features.

[sambuca]

I don't get why it have to be so hard, is there really any reason LinuxMCE actually needs to control your network and determine the design of it? The only reason I can see is PXE, but any DHCP server should be able to handle that. I do modern multi tier software for a living, and usually it's just a matter of configuring a few addresses so that the parts find each other.

The more I look it, the more I think that some other solution might fit my needs better, such as XBMC. Sure, it will take a bit of work to integrate my alarm sensors, but that's a later project anyway. At least that's just a simple install, and no need to rebuild/rethink my network. Too bad, LinuxMCE really has some nice features.

I think the main issues have been mention in this thread, but just to make it perfectly clear:
* Having LMCE run DHCP will make devices Plug and play (like new Media Directors, Squeezeboxes, and many more)
* Not running DHCP on LMCE means you have to set up these devices manually yourself, and of course do some setup of PXE/DHCP. Don't know if this is documented anywhere, besides in the code itself, so you might be on your own here.

Whether you use the firewall/gateway features of LMCE or not, is not really an issue, imho.

So while there definitely is a reason, you should be able to make both options work. But as this falls outside the supported scope of LMCE, my guess is that you won't get much help from the (already too busy) developers.

best regards,
sambuca

[wierdbeard65]

It would be a shame if you didn't proceed with MCE. You obviously have extensive networking experience which would be valuable to the community and if you have a Linux programming background as well, then so much the better!

A lot of the problem here is that (I guess) the majority of the current users are tech-heads who don't mind diving in and playing. This must be so at the moment as the system is not yet "plug and play" (more "play and pray"!). The downside is that because we think we know what we are doing, we tend to over-think it and wnat to change the setup. The target, however (at least as I understand it) is for an appliance. Simply plug it in and you have a media-aware, home automation system that can manage your network. Something that the "man / woman on the street" can use. Just like buying a DVD player now.

FWIW, I'm seriously considering going down the "DHCP but not Router" route myself. The problem I face is the SWMBO needs the internet to work. All the time. No excuses. With MCE being beta, every time an upgrade / update happens, there is always the risk that it will break and SWMBO falls off the 'net. If that happens, project cancelled. No appeals. No second chances.

[Troberg]

Don't get me wrong here, I love the scope and ambition of the LinuxMCE project. I'd love to help, but I'm not a Linux programmer (yet, I will eventually, but I'm still looking for a nice RAD environment), and at the moment, I have other concerns which fill my time.

The way I see it, LinuxMCE is the by far most ambitious project of its kind, taking it beyond a simple media center into a home center. The modularity and configurability is very imressive indeed.

What I feel, however, is that any single program that wants to control my network is a problem. One such program can probably work nicely, but what happens when the next such program comes along? That's the reason I try to maintain a strict "one program does one thing, and does it good (and that goes double for server software)" and a "each machine only carries a single server service". That way, there's a minimum of collisions, and if a machine dies, only a single function on the network dies, and it will be simpler to replace. It also allows me to replace services one at the time, which minimize the risk of everything becoming very confused.

Also, by trying to do things outside the main scope of the product, such as providing basic network infrastructure, chances are that those odd bits will not be as full featured, as well done and as well maintained as a dedicated program. Just look at Outlook, which I'm forced to use at work. Instead of being a mail client, it's also a customer database, calendar, booking tool, and does neither good. If they had focused their efforts, they could have made a decent mail client, now, instead, it's a crappy mail client, crappy customer database, crappy calendar and a crappy booking tool.

In this case, it does provide some neat features, such as PXE, but it's a bit scary to replace such a vital part of a complex network as the DHCP server. Chances are that there will be problems, and with them, downtime.

[wierdbeard65]

Don't get me wrong here, I love the scope and ambition of the LinuxMCE project. I'd love to help, but I'm not a Linux programmer (yet, I will eventually, but I'm still looking for a nice RAD environment), and at the moment, I have other concerns which fill my time.

I thought you said you were a programmer? RAD??? Bah! LOL 

What I feel, however, is that any single program that wants to control my network is a problem. One such program can probably work nicely, but what happens when the next such program comes along? That's the reason I try to maintain a strict "one program does one thing, and does it good (and that goes double for server software)" and a "each machine only carries a single server service". That way, there's a minimum of collisions, and if a machine dies, only a single function on the network dies, and it will be simpler to replace. It also allows me to replace services one at the time, which minimize the risk of everything becoming very confused.

I agree, but then you are a tech-head. MOST home users don't have the money / space / power / experience / need to do that. A simple "box" that sets up a complete media-aware network with home automation is what's needed That's why products like BT's "Home Hub" in the UK are so popular. Indeed, it's why most homes have a combined ADSL router / switch / WiFi Access point. (In fact, there are even techies who now confuse the three and refer to, for example, the access point as a router. - Flame on!)

Also, by trying to do things outside the main scope of the product, such as providing basic network infrastructure, chances are that those odd bits will not be as full featured, as well done and as well maintained as a dedicated program. Just look at Outlook, which I'm forced to use at work. Instead of being a mail client, it's also a customer database, calendar, booking tool, and does neither good. If they had focused their efforts, they could have made a decent mail client, now, instead, it's a crappy mail client, crappy customer database, crappy calendar and a crappy booking tool.

There's an old saying "Jack of all trades, master of none" which sums that up. Couldn't agree more. However, be careful aboiut what is, or isn't part of the main scope. DHCP is how pnp works, so it's part of the scope

In this case, it does provide some neat features, such as PXE, but it's a bit scary to replace such a vital part of a complex network as the DHCP server. Chances are that there will be problems, and with them, downtime.

Like I said above, you are not a typical target user. Think of the home owner who doen't have an interest in IT as such. MCE is supposed to be an appliance

[trentend]

Don't get me wrong here, I love the scope and ambition of the LinuxMCE project. I'd love to help, but I'm not a Linux programmer (yet, I will eventually, but I'm still looking for a nice RAD environment), and at the moment, I have other concerns which fill my time.

The way I see it, LinuxMCE is the by far most ambitious project of its kind, taking it beyond a simple media center into a home center. The modularity and configurability is very imressive indeed.

What I feel, however, is that any single program that wants to control my network is a problem. One such program can probably work nicely, but what happens when the next such program comes along? That's the reason I try to maintain a strict "one program does one thing, and does it good (and that goes double for server software)" and a "each machine only carries a single server service". That way, there's a minimum of collisions, and if a machine dies, only a single function on the network dies, and it will be simpler to replace. It also allows me to replace services one at the time, which minimize the risk of everything becoming very confused.

Also, by trying to do things outside the main scope of the product, such as providing basic network infrastructure, chances are that those odd bits will not be as full featured, as well done and as well maintained as a dedicated program. Just look at Outlook, which I'm forced to use at work. Instead of being a mail client, it's also a customer database, calendar, booking tool, and does neither good. If they had focused their efforts, they could have made a decent mail client, now, instead, it's a crappy mail client, crappy customer database, crappy calendar and a crappy booking tool.

In this case, it does provide some neat features, such as PXE, but it's a bit scary to replace such a vital part of a complex network as the DHCP server. Chances are that there will be problems, and with them, downtime.

As far as I can see you're really missing the point here. Linuxmce is designed to work by taking over DCHP and allowing the media/automation network to be plug and play.  It is designed for adoption in non-technical home networks (although it may not be suitably easy to setup for that role yet....) to work in an exclusive mode on the network.

If you have sufficient network knowledge and skill (not really that much), it is possible to run it on a separate exclusive subnet, which in itself is a relatively simple networking problem.

That's it.  It's that simple.  Set up a subnet and you can use Linuxmce as intended.  You my use lots of words to describe how big smart and efficient your network is.  As you have described it I feel that I could achieve what you want to. It's simple.  S-U-B-N-E-T.  You will almost certainly need at least one managed switch.  This is a consequence off having a complex home networking requirement outside the scope of almost any equivalent project.

If you don't want to do it that way, then that's fine, but you'll probably struggle.  I'm no expert, but I really think it's that simple.

[Troberg]

I thought you said you were a programmer? RAD??? Bah! LOL

Well, at work, I get paid by the hour, so it doesn't matter. At home, I want to get things done, then have some spare time so I can get other things done.

I agree, but then you are a tech-head. MOST home users don't have the money / space / power / experience / need to do that.

Well, some people claim that I'm not an average user. I claim that I have so many machines that the average is skewed my way (if I use 25 machines, shouldn't that count as 25 users in the statistics).

As far as I can see you're really missing the point here. Linuxmce is designed to work by taking over DCHP and allowing the media/automation network to be plug and play.

No, I get the point, it's just that my needs are different, which means that LinuxMCE may not be for me. I'll do a small scale test with a few machines on a separate network this weekend and see what I can do with it, but for now, my gut instinct tells me that LinuxMCE fiddles around in too much stuff that I don't want fiddled.

I'm not saying that the concept is wrong, I'm saying that I might not be the target for such a concept. A less intrusive system which is just one application among others and no ambitions to become "supreme ruler of my network" , such as XBMC might be a better fit for me.

That's it.  It's that simple.  Set up a subnet and you can use Linuxmce as intended.  You my use lots of words to describe how big smart and efficient your network is.  As you have described it I feel that I could achieve what you want to. It's simple.  S-U-B-N-E-T.  You will almost certainly need at least one managed switch.

No, it would require all my switches to be managed, as the media players is placed more or less at the far end. As I said earlier, I've had problems with managed switches, and I'm not prepared to go down that road again. I want the intelligence on my network to be in the computers, not in the switches. Also, I want to be able to reach every machine from every other machine, if nothing else because it makes trouble shooting so much easier. I also suspect things will be easier if the file servers are on the same network as the media players, but I also need my desktop machines and laptops on that network, and they should not receive DHCP from LinuxMCE.

Also, if one starts designing the network to accommodate specific software, I'll probably run into two problems down the line:

* Some other software wants it some other way.
* Some update will require me to rebuild my network.

Also, I'm lazy. I don't want to reconfigure the network more than absolutely necessary. I have lots of things that will start to break, such as database connections and such stuff, if I start dividing the network into separate subnets. Sure, I can fix it, but such tasks have a tendency to snowball into an avalanche of related issues, until I get enough and just goes back to the way it was.

[skeptic]

I/we also don't fully understand your setup.  I.e. which servers are internal vs. external, what spof path are you concerned with,

There is no clear distinction between internal and external servers. The servers which are accessible from the outside also have an internal NIC, which makes them a lot easier to manage. As I have them pretty much locked down and behind strict firewalls, I don't see this as a major security breach. In other words, they have two connections, one to the firewalls, one to the LAN.

A diagram would be really helpful.

Your suggestion would create a bottleneck between servers and desktops which would be a spof.

Nope, it wouldn't. 

What I want/need is a way to have a nice integrated media solution without having to rebuild/rethink my network, at least not in a major way. I don't get why it have to be so hard, is there really any reason LinuxMCE actually needs to control your network and determine the design of it? The only reason I can see is PXE, but any DHCP server should be able to handle that. I do modern multi tier software for a living, and usually it's just a matter of configuring a few addresses so that the parts find each other.

You can have a nice integrated media solution without rebuilding/rethinking your network.  The only big changes would be static IPs for your servers (if you haven't done that already), and changing to 192.168.80.x IPs for those devices you want on the LMCE network (same physical network I realize, but also same subnet to avoid traffic going through a router and hitting the physical network twice).

The more I look it, the more I think that some other solution might fit my needs better, such as XBMC. Sure, it will take a bit of work to integrate my alarm sensors, but that's a later project anyway. At least that's just a simple install, and no need to rebuild/rethink my network. Too bad, LinuxMCE really has some nice features.

Honestly, I think you are struggling too much and making it harder than it is.

[Troberg]

A diagram would be really helpful.

Yep, I know, but I don't have any that's up to date since I moved.

Nope, it wouldn't.

Am I misunderstanding what you say? I interpreted it as splitting the network in two subnets, putting the LinuxMCE server between the "server network" and the "client network". That would be a spof.

You can have a nice integrated media solution without rebuilding/rethinking your network.  The only big changes would be static IPs for your servers (if you haven't done that already), and changing to 192.168.80.x IPs for those devices you want on the LinuxMCE network (same physical network I realize, but also same subnet to avoid traffic going through a router and hitting the physical network twice).

The servers are mostly on fixed IP's today, as is some development machines (easier to debug network code with fixed IP). Some servers are not (mostly the download slaves), but could be. It would be good if these can remain on the same IP's and on the same network, as there are lots of "hidden" references in configuration files for various programs.

For practical reasons, I also want my desktops and laptops to be on that subnet.

What would the gain be in having the LinuxMCE machines on a different network, especially as the big file servers will remain on the network with servers and desktops/laptops?

Surely, it can't be that hard to put LinuxMCE on the same subnet? In my experience, if you don't have any compelling reason to split into subnets, don't.