Re: The Case for a Pocketable Server
 

Info about new Atom dual-core

(There are many x86 game servers available for Linux.)

In that case the architecture doesn't matter much although I find OpenblockS resources:

Limited. This is half as much RAM as the NIT has, and the processor seems slow, too, athough it is a RISC.

Its more meant as router. 500 USD for such a DIY router is quite laughable though, at least IMO. Especially if you consider it only has 2 10/100 LAN ports.

I think you'll have most luck with a thin client or desktop although these don't have a LCD touch screen by default.

An example of a thin client now. CherryPal has this:

And VGA, and a lot more. Small, uses 2W.

:o my mistake calling it CherryPC that must have been a Brain glitch. Damn you, Pinky.

Re: The Case for a Pocketable Server
 

I read here it uses 4-6W but that must be excluding the HDD? Flash with PATA interface or SSD (40 or 44 pin?) is an option though.

Re: The Case for a Pocketable Server
 

What development? Drop Debian, Ubuntu or Ångstrom (for the best armv7 compatibility) on there and be done with it. All three have good arm branches, so there's really no benefit to x86 here (besides, we're not dealing with binary-only software anyway).

And what cost? The Beagle is $149 out the door. Certainly sounds cheaper than the FitPC. Drop in the rest of the parts to turn it into a server and . . . you're about on-par with the FitPC. Heck, the overo sounds like it may be even cheaper (and will bundle additional stuff like Bluetooth and WiFi).

I do have a bias towards OMAP and ARM. x86 is a commodity. It's not interesting—"Oh, boy, another cheap x86 computer . . . haven't seen that before." Building an x86 something or other is old news. It's been done to death and nobody really cares anymore.

ARM on the other hand. In this in this sort of performance bracket? It's is new and exciting. It hasn't been done to death. It's fast, low power, and interesting.

Besides, since it's a pocketable server that seems to be being outlined here, it's gonna run off a battery and x86 is a major handicap for that sort of usage.

For the record, x86 simply can't compete with ARM power-wise. Looking at Wikipedia, the Geode takes up almost as much juice as the whole Beagle Board (1.8W for the Geode vs. 2W under load for the whole Beagle).

Re: The Case for a Pocketable Server
 

None of the above really makes a case for nor against anything at all. If what we want is a pocketable/backpackable server, who cares whether it's commodity vs geeky-chic? What matters is: does it get the job done in a small enough form factor?

THAT makes the case against x86.


The fitpc wants 12V and 4-6W. What size battery does that take if you want it to last 4-7 hours (ie. comparable to the NIT)? Does anyone pre-make a battery that could easily be connected to the fitpc?

Are there any pre-packaged solutions for the beagleboard? What are their power and battery requirements? (if the idea is to build something from boards, why not just build peripherals for the BUGbase?)

Anyone have an idea about the similar requirements for the BUGbase? (I couldn't find specifics about power draw nor battery lifetime; it has internal batteries and support for external batteries)

Re: The Case for a Pocketable Server
 

I care, and it's only my personal opinion that I'm elaborating on.

Not currently, but the Beagle guys would love for you to build one. ;) 2W at 5v is about the max you'll ever use. Basically, the same power requirements as the tablets minus the screen.

Because the Bug uses a slow ARM11 and costs $350 just for the base.

Re: The Case for a Pocketable Server
 

I can see the price being objectionable, but I don't care about the slower speed. How fast does a router and file server really need to be? (though, I guess if you're going to use it for a game server as well, that changes things; but that's not part of my needs for such a device)

Re: The Case for a Pocketable Server
 

Depends on the purpose. Check the price of current consumer & SOHO grade routers and NAS. The NAS usually have 100 mbit, and they choke on throughput, or they're expensive. Look up Synology and Thecus.

Texrat's FP does state he's interested in a game server so resources are something to keep in mind IMO.

People do care about making a x86 device when x86 architecture makes sense. Usually, thats not the embedded market though, minus the higher end potential of Atom. (I do consider a nettop or MID embedded.) Most people don't give a rat about architecture zealotry...

Re: The Case for a Pocketable Server
 

I'm not sure how or where that comment fits into what I was discussing...?

Re: The Case for a Pocketable Server
 

Because some people who'd like to run a game server would want this functionality. You not, so you can ignore it :confused:. We all have different needs. E.g. some care for the architecture used for the project.

Re: The Case for a Pocketable Server
 

Ah. Okay. Let me qualify an earlier statement: in this thread, I'm open to all sorts of ideas that don't unfairly restrict the openness and portability requirements. Those are fundamental.

So hosting/lobbies need to rely on completely ubiquitous standards like html, css and soap. Apps should use one of the portable executables I mentioned (I am hopeful for java, eventually).

Re: The Case for a Pocketable Server
 

Ahhhh, I'm finally starting to comprehend your goal, and some of its potention. You really do want a a kind of 'mobile LAN party'.

One could add games to the server, and have it e.g. in the car, so the kids can play a game while you're on the road.

Tetrinet might be fun for this. Tetrinet Java server.

For Silverlight you can try Mono + Moonlight, but when I tried to use it to play WMV it didn't work, and I don't know how advanced these technologies are nowadays.

For Flash, the device only has to host the Flash file, not support it.

Have you waded through Linuxdevices.com archives?

Re: The Case for a Pocketable Server
 

To me it means games with open-source servers; there's an old zDaemon codebase that was open-source, and Q[1-3] at least are completely open software.
So I'd say, it needs to be UNIX, as the (non-interpreted) open-source game-servers I'm aware of all come in UNIXy variants. Therefore Linux, out of popularity, hence hardware support.

I'm definitely gonna say OMAP3 is the best option this way I've seen; the various x86s all lack efficiency, performance, or both.
I'm inclined to say one full-SD is all it needs built-in if it has appropriate interfaces for expansion.
n would be nice here, to get best range for uplink purposes, but probably g...Probably need to push this offboard.
Including zealous autoconfig, of course.
Do we need slave mode? It doesn't hurt. USB networking or serial for another internet-access channel, and since OMAP35xx all have them, it's cool.
Possible, maybe, but I'm afraid pretty hard for a start-up operation. (Or are you trying to get Nokia to do this? ;)) I can see going to $400.
Much as I'm tempted to say yes, just HTML, that's not really cross-platform enough. Dual interfaces via tty and HTML, with the tty access possible by USB/BT serial or telnet. Seriously, it's only software, and if content and presentation are properly segregated in design, there's no reason C64s, Newtons, and anything else lacking a modern browser shouldn't be able to hookup for a little fun.
Practically necessary. And miniSD WILL NOT CUT IT!
-1U max size[/quote]What? not 1U of 19" rack, I hope! It should be thinner, anyway.[quote]
Yes, but more than one; USB is our main expansion option.
Hot-swappable implies a built-in battery separate from the main battery. To me, this means a base unit with a small battery (10-15 minutes runtime, maybe an hour) that will be recharged from an external DC source. This external DC source may be either a big (10+ hour) backpack, or a wall-wart/car adapter/etc. To leave charging hardware out of the device, I'd tend to make the wall connector multipole, so it provides power to the device and a separate connection to recharge the battery.
This gets bulky, and there are at least three sane alternatives: 1.8" (whether SSD or HDD), CF/MicroDrives, and USB drives. They should be modular, and you can add any one (or more) of these.

The SD RAID solutions I'm aware of are 2.5" SATA, BTW, so support for them is best accomplished with a 2.5" drive adapter.

Maybe close in some ways, but 3A is a lot of juice. This thing might have the power needed, but it's a little big (thick, mainly), not modular enough, and eats way too much power. IMHO, of course.
All good modular things, but not in the core.
This is a software-only problem, assuming we've got USB ports, and that's how you plan to connect the printer. And the software needed is completely in existence. It should definitely be included.
External screen capability is really important. It won't be used much for control, but if you can get a projector (say, sneak into a unused classroom between classes...), some good live scoreboards can be shown. Again assuming OMAP35xx, we've got a video out interface; it would be foolish not to have a connector.

It looks like USB can work for the modules except the power-pack. The others might need power connections beyond the 5V/500mA that could be properly provided by USB, so that might have to be provided as well, but I hope not.

So I'd say either a Beagleboard or a Gumstix Overo (from what I've seen) for the main board. The main module is almost the footprint of a 3.5" drive, but thinner. Maybe half as thick, if possible. It has connectors all around, and something like 2 or 3 AA worth of battery.

• HDMI (DVI is too big, use a dongle!)
• USB OTG micro-AB (mini-B's not standard, and micro-A cables are starting to exist...)
• USB ports for expansion. Make these full-A, and stackable modules can just have short pigtails to plug here. I'd say 4.
• BT is internal, with built-in antenna.
• WiFi is internal with antenna but has (room for, at least) external antenna connector.
• 1 external and at least 1 (hopefully 2) internal SD. One internal SD might be microSD, e.g. the gumstix board... (It seems from a brief glance this has one microSD and runs one mmc interface to the connector; if that's right, then one is wasted...)
• Audio input/outputs.

A battery module can clip on the bottom or top; I originally considered a battery module on the bottom and other modules on the top, but if pigtail interconnects (rather than PC/104esque passthroughs) are used everything can stack in any order.
Other modules include:[list][*]PCMCIA -- one type III Cardbus slot; with an electromechanical adapter, CompactFlash/MicroDrives. Good for storage, ethernet and exotic comms stuff. (Ethernet might make sense here, if you have an old 10/100 card laying around.) Maybe a thin spyder variant would take Type II cards while stacked anywhere, or a Type III sticking out the top, if it was on the top; it would still stack and function in any postion.[*]USB-only ExpressCard/54 -- this will work with EC/34 or EC/54 cards, for 3G, storage, ethernet, etc. (This is an electro-mechanical adapter + a voltage regulator.)[*]PATA and/or SATA -- 2.5" drive adapter, could also be used (with spacer, or just additional screws? IDK.) for 1.8". Maybe it's worth making one for 1.8" only; that might fit two 1.8" in one (thinner) case.[*]Anything else running off a USB connection could be made into a module. Some of these make more sense than others, naturally; an optical mouse module would be just wierd, while a GPS module would probably make sense.

These modules all have some sort of keyhole tabs (top and bottom on ALL units) to click together rapidly, and grooves in opposing sides to permit tie-wrapping a stack together for semi-permanent mechanical assembly. Modules have no standard height requirement, and should be as thin as possible without (inordinately) compromising functionality. There are no top-only or bottom-only devices (though modules may gain capability in a particular slot), and pigtails should be just long enough for any reasonable stack height (given a maximum of 6 bricks in a stack: core, battery, and 4 peripherals).

Modules need not be standard USB, of course; as the design as written has no speakers, maybe a speaker module with a 1/8" TRS pigtail is in order? As long as the mechanical interface is simple and generic, and users are prepared to deal with the consequences (carefully ordering stacks to avoid running out of pigtail length!), anything goes. If you wanted, for convenience, you could make the wall-wart stackable, too; again, it has click-tabs top and bottom, so it can be added to either end of a stack, bringing it pretty close to cube form, but ready to travel.

Normally, you'd only need the core and battery modules, which would fit in normal pants pockets. With an expresscard module and a 3G adapter (reasonable even if it's accessed by an N9xx, because you don't want one client dealing with all the traffic from all the clients), it might be a tight squeeze, but doable; definitely in a hip pocket, and perhaps pulling it out to sit down. The "full stack" configuration, of course, would vary depending on what you filled it with; a 2.5" HDD-based jukebox with a speaker module would come out taller than one with 4 one-slot MMC readers.

Software elements in next post. Stupid limit...

Re: The Case for a Pocketable Server
 

Nah, just wanted to start freewheelin'... :D

Re: The Case for a Pocketable Server
 

Re: The Case for a Pocketable Server
 

Thanks Benson. I only have minor quibbles and I'll elaborate later.

Re: The Case for a Pocketable Server
 

TI has a pico projector coming out just for the Beagle, this might be useful here.

Barring really crazy price differences, I think the question comes down to whether or not you want the PowerVR and DSP (overo has neither). The pico projector with scoreboard presents an interesting opportunity to have the server machine spectating and showing different players on the "big screen". That'd be with the Beagle and the PowerVR, of course. ;)

Re: The Case for a Pocketable Server
 

Yeah, that's what I thought I was seeing; I hadn't thought about anything beyond scoreboards. I couldn't see any pricing info on the Overo (at a glance), but figured it would probably be cheaper in light of that difference, hence a better pick. But accelerated graphics for chasecams, etc. could be quite nice.

Re: The Case for a Pocketable Server
 

Yeah, I'm not sure. My gut tells me that the cheaper OMAP3503 could be offset by the Bluetooth and WiFi bundled, but who knows. I guess we'll see what happens come November (Gumstix is waiting on ES3.0).

Note: overo is properly not capitalized. Much like Maemo used to be. ;)

Re: The Case for a Pocketable Server
 

Well, either the folks at gumstix.net have pulled a Maemo and changed overo to Overo, or someone needs to jump on their webmaster; http://www.gumstix.net/Overo/ and below is full of Overos, so I was led astray. I actually had noticed some news articles calling it overo, but thought they must be wrong or old. :o

Re: The Case for a Pocketable Server
 

Hrm, interesting.

Well, then, I'll stay out of it. ;)

Re: The Case for a Pocketable Server
 

Assuming that they ever actually release this thing... what about the CherryPal?

5.8" x 4.2" x 1.3" in 10 oz.

2W

400 MHz Freescale, VGA, 10/100 Ethernet, Wifi, 256 RAM, 4GB Flash, 2x USB

Debian

$250

Probably not very good as a game server ... but attaching various USB things to it (card readers, WWAN dongles, etc.), and putting a battery on its power input, might make it a good storage/gateway type backpackable.

http://www.cherrypal.com/products/

Re: The Case for a Pocketable Server
 

Yes, I referred to it earlier. You can probably pick a different OS if you'd prefer. I doubt the 2W though, and I'm 100% sure you won't have 2W after you attached a HDD. A HDD uses approx 30W during spin up (hence staggered spin up is important sometimes), and approx 9W when idle. You'd have to opt for a flash based (SSD or not) PATA device, not sure if it'd require 40 or 44 pin though. For this purpose a 4 or 8 GB is probably enough. Those cost about 50-100 EUR here.

Re: The Case for a Pocketable Server
 

I wouldn't go with an HDD. I'd go with some sort of SSD/Flash based storage. There's DIY SDD adapters that let you put a CF card into a 1.8" drive slot, or 6 SD cards into a 1.8" drive slot. I'd go with one of those. Though, I'm not sure what's currently a good MB/s rate for that kind of thing (I want lower power than the HDD, but still at least HDD speeds, if not faster ... you know, the holy grail of SSDs ;-) ). But, you can get a CF card -> 1.8" drive adapter for $30ish, and a 32GB CF card for $100ish.

I've seen an adapter that says CF cards get up to 40MB/s, and I saw an SSD that does 36ish. Not sure what's normal for a laptop/iPod SSD though.

For the CherryPal though, you wouldn't use the 1.8" drive adapter, you'd just have a USB CF card reader. That should be a bit cheaper than $30. An ideal companion to something like the CherryPal would be a device that had an internal battery for powering itself and the CherryPal, so one connector to the cherrypal's power port ... and then also had a connector into the CherryPal's USB ports that provided 2 SDHC card slots, a CF card slot, and an ExpressCard slot.

Though, I also wish the CherryPal had a DVI (with DVI-A) interface instead of VGA. VGA is like having a PCMCIA card slot these days ;-)


As for OS ... Debian is MORE than acceptable for an in-backpack server. It doesn't need the pretty UI of Ubuntu or Maemo. I probably wouldn't even install X on it, if i didn't have to.

Re: The Case for a Pocketable Server
 

If you meet a DVI-only device you can use a converter. VGA <-> DVI converters exist.

I'm not sure if I/O is important for Texrat, or in general. If not, even a USB stick might be a possibility, but it might not be elegant having something 'stick out of the device'. Don't forget these have different speeds, too. Some good webshops provide the speeds in their inventory list.

SSD is said to be slow on many small files. There are also now faster SSD with SATA2 support. As you can see from the list hereunder SSD doesn't have to be very expensive.

http://tweakers.net/pricewatch/cat/674 (I have a PATA Transcend)

http://tweakers.net/pricewatch/cat/48 (tons of several options)

Debian is good, especially as default OS, but one might prefer a different OS.

EDIT: I would swear I saw embedded devices with DVI (e.g. on picco.nl), but x86 (VIA for sure (Jetway mainbords), and perhaps AMD Geode). Not any RISC-based platform.

Re: The Case for a Pocketable Server
 

I have yet to see a VGA -> DVI converter... (what your two way arrow implies). It's not cheap to do that because you need to do an A2D conversion.

But yes, the fact that DVI-A -> VGA converters exist is why I think every device that's going to have an external monitor jack should have a (micro|mini)?DVI jack with DVI-A support on it. It can be converted to VGA (DVI-A is the analog part of the DVI spec), and to HDMI (from the digital part of the DVI spec).

DVI with DVI-A is the "universal donor" of monitor connectors.

Re: The Case for a Pocketable Server
 

Hmm not sure we understand each other.

Device with VGA input (female)
Device with DVI output (female)

Then you need a DVI (male) to VGA (female) converter on the 2nd device, and use a normal VGA cable. You normally carry your VGA cable with you (or so I'd assume) having your converter with you in case you stumble upon a DVI only output.

If the second device has for some reason DVI male output you'd need a slightly different converter: DVI (female) to VGA (female)

Re: The Case for a Pocketable Server
 

Two gender changers and DVI -> VGA should work fine... No, I'm not being serious here, although no doubt there is some Chinese factory, somewhere, turning out DVI gender changers for no evident reason.

No reason VGA -> DVI-A wouldn't work, but from my limited experience/knowledge of displays, I think ones with both RGB and DVI interfaces generally have a VGA port and a DVI-D, rather than a single DVI-I.

(For reference, DVI-D is DVI with only digital signals, DVI-A is DVI with only analog, and DVI-I is Integrated, i.e. both, the universal donor you mentioned.)

Re: The Case for a Pocketable Server
 

Well, I'll give an example. I have a KVM with DVI. Its output is 2 DVI male. On my workstation it works (on DVI female), but I'll need a DVI male -> VGA female in order for it to work on my laptop. My graphics card came with VGA male -> DVI female which is exactly the other way around (it is what you'd need though). I haven't searched around yet on local stores, but I did see it on eBay.

Re: The Case for a Pocketable Server
 

Interesting discussion. I'm partway there, running Nokia's Mobile Web Server on my N95. Granted, I cannot develop and therefore don't have much more than a few web apps running on it; but its avariant of Apache and so I'm sure that if its there that its possible to do and modify for the IT's hardware.

Re: The Case for a Pocketable Server
 

I was actually just wondering about the feasibility of using an N95 as a pocketable server. Esp. if you could put Android (or some other linux variant) on it. Not sure. Or, if Android wont run on it, there's always putting Android on one of the kaiser family of phones, or something.

Re: The Case for a Pocketable Server
 

That's actually a cool idea on its own, although the cost is a bit high. You would have a built-in modem...

Re: The Case for a Pocketable Server
 

I was going to avoid the cost side of it by buying one used on ebay.

Though, with this news bit:

http://linuxuk.org/node/65

I had an idea wondering how hard it'd be to:

• add USB Mobile Modem support to the N800 (or N810)
• add an "access point" mode to them (either via the internal wifi, or via an external USB wifi dongle)
• add a mode for acting as a Bluetooth DUN _host_ (instead of device) (maybe PAN as well), if you can't do the wifi access point
• add a light-weight SIP server to it, that can use a USB modem for voice calls
• add a light-weight Jabber server that can use a USB modem for SMS/MMS
• possibly write the above things so that you can attach multiple USB modems (in case you want to do voice/SMS/MMS on one, and data on a different one, or have multiple data feeds, or multiple voice lines, etc.)
• hopefully write it such that you can do voice/messaging/data simultaneously on the same modem
• Dunno if you can or can't do USB storage devices now, but make sure that's in there
• All of the above with friendly configuration interfaces -- maybe not Maemo Control Panels, but if not then a webmin server that could run on the tablet.

That would probably handle everything I want/need in a pocketable server. I might want to also buy a battery powered USB hub, and stuff it all into a semi-large Pelican Case or Otterbox, but, here's what I'd see as the strengths:

• I already own 2 N800s and an N810
• I could use one as a network gateway device, with any USB Modem bearing carrier, or multiple USB modem bearing carriers, and serve it to one of my other tablets or even to a laptop (or wifi enabled phone that may or may not have 3G itself)
• With the N800's two built-in SD slots (can they support the new 16GB SDHC cards?), the possibility of USB card readers (CF, SDHC, etc.), and possibly even USB HDs (HDD or SSD), you'd have a ton of storage options
• With the software, you'd be able to use any IM client for your texting, and any SIP client for your voice calls. Possibly even the same tablet itself (the RTCOMM software on the Maemo talking to the local SIP/Jabber server on the same tablet).

I actually rather like this idea... throw in SyncML client (for Maemo in general) and server (for use as a pocketable server) and an external battery pack (I'm pretty sure those already exist for devices that use the N800/N810/E61/E62/E71/etc. style charger plug), and it'd be gold. A battery pack that could both recharge the N800/N810's battery, and power a small USB hub, would be ideal.

Of course, you might be able to do the same things on a Pandora, I'm sure. That way you don't have to worry about the EOL of the N800.

Re: The Case for a Pocketable Server
 

Or, someone could just take Raccoon, which is the "mobile" Apache-like server that the S60 Mobile Web Server uses, and build from there since most of the main parts are done already and its already open source.

All that's left is to supply a tablet that can do WiFi and cellular connectivity...and according to M5 that's coming. tada

Re: The Case for a Pocketable Server
 

I guess since everyone's talking about it in other threads, I probably don't need to mention that you could look at the Pandora as the kind of server you're talking about?

Re: The Case for a Pocketable Server
 

Yes, but it (and the tablets, and the next tablet) is a little weak on expandability; if you modify the base with a couple notches for attachments, and keep it as top board, it would serve nicely.