Of course, if anyone wants to pay the monthly bills for the current system, that works too.

Keeping the current system running costs between $4300 and $5500/month depending upon load.

Here's the breakdown for December 2012:

Amazon CloudFront $18.02
   US $17.15
   Europe $0.76
   Asia (Tokyo) $0.05
   Asia (Singapore) $0.02
   South America $0.02
   Australia $0.02
Amazon Elastic Computer Cloud $2260.10
  US East (Northern VA) $2100.31
    EC2 Linux
      Standard Small 744 hours $48.36
      Standard Large 2232 hours $580.32
      Standard Extra Large 2232 hours $1160.64
      Micro 2976 hours, 59.52
    EC2 EBS
      969 GB-Mo $96.90
      25,554,572 I/Os $2.56
      snapshost 94.857 GB-Mo $9.01
    Elastic IP
      1485 hrs $7.43
    Elastic Load Balancing
      685.667 GB $5.49
      5203 hours $130.08
  US West (Northern CA) $68.21
    EC2 EBS
      245 GB-Mo $26.95
      snapshot 17.290 GB-Mo $1.82
    Elastic IP
      3719 hours $18.60
    Elastic Load Balancing
      744 hours $20.83
  Asia Pacific (Singapore) $91.53
    EC2 EBS
      523 GB-Mo $57.53
      snapshot 319.222 GB-Mo $30.33
    Elastic IP
     743 hours $3.72
Simple Storage Service $535.93
   US Standard Region $524.89
     first 1024 GB-Mo $97.28
     next 4012.338 GB-Mo $320.99
     2473612 PUT COPY POST LIST requests $24.74
     81887553 GET request $81.88
   Asia Pacific (Singapore) Region $11.04
     116.105 GB-Mo requests $11.03
     3 GET requests $0.01
Amazon RDS $991.66
   US East (Northern VA)
     RDS storage
       provisioned storage multi-AZ 55.000 GB-Mo $11.00
       270577727 I/Os $27.06
       provisioned storage 50.000GB-Mo $5.00
    RDS for MySQL
      RDS large 744 hours $271.56
      RDS large multi-AZ 744 hours $543.12
      RDS small 744 hours $133.92
AWS Data Transfer
    US East (Northern VA) $637.57
      data transfer out 5105.833 GB $612.71
      regional data transfer 2286.435 GB $22.86

Total $4441.28

Well, Zurk's firmware relies on you getting the widgets from *somewhere* initially and storing it on a dongle, and it costs something to host and serve all of those widgets.  They're actually stored in three places - when you upload a widget, it's stored as a blob in the main database (one of the large RDS instances), then duplicated into the database slave (the other large RDS instance).  There's a script that runs periodically that then copies the widget to S3, where it's cheaper and faster to serve them.

Back when Chumby was operating, it actually cost somewhere north of $25K/mo to run the service - we were required to run it on actual hardware in an actual datacenter to comply with rather strict data security requirements of one of our partners.  Once we severed that relationship, we moved to Amazon and I modified the Control Panels to be much more aggressive about caching, which took us down below $6K/mo.  Amazon has been steadily reducing prices so we're now where we are.

There's a lot of data stored - I think even if we suspend all of the EC2 and RDS instances, the storage of data alone runs about $900/month.

Some of that data include the entire subversion history of the company - many things could be eliminated or taken offline.

There are currently just around 100,000 devices registered in the system since the Sony dashes were moved out.  Of those, about 40,000 are in active use.

No, Zurk's firmware comes with a selection of widgets and serves them to the control panel off the Flash drive on which you install his distribution. You can modify the channels he provides and add new widgets to them (even add new channels if you like) as long as you have a copy of all the files needed to run each widget. The problem is that such editing is not particularly easy; it involves modifying an XML file, and it's easy to make a mistake in doing that kind of thing. I don't think it would be too hard to write some kind of Perl script to do most of the editing offline, but from what Duane says it's likely to be pretty costly to keep the existing web environment.

Duane, what do you think of the idea of trying to at least keep the current widget files available for manual download, to use with something like Zurk's firmware? Feasible? I would think only someone who owns the Chumby rights could do that legally though.

I love my Chumby alarm clock, but almost $5,000 a month is some serious cash to find, I wish you luck!

I too am willing to donate. I am not rich but I would like to help. $5,000 a month is a tough nut to crack. Even if we get enough people to donate this month, will it keep up month after month?

I wish I could be of more help (coding / financial). Duane has done a magnificent job and we are all in his debt. If this forum goes dark is there someplace else we can send/receive messages concerning all things Chumby?

If I build a replacement system, it will not cost $5K/month.  My guess it will be in the low hundreds - my goal is under $200/month.

What you're seeing with the current system is a "professional" system, with full analytics, the complete build system, monitoring, backups, the whole source control history back to day one, system logs, internal files and documents, etc even though in some cases the servers that server this stuff have been shut down.

At $4400/month, with 40K active units in the field, it's actually only $0.11 per unit per month, which is pretty low.

Some of the costs are independent of the number of units - the servers have an hourly cost, and the storage is used whether or not anyone actually fetches the data.  The variable costs are the amount of data incoming and outgoing on the system, which *is* dependent upon the number of units.

If you look at the costs posted above, the variable costs are toward the bottom (the data transfer), the I/O to'from the RDS databases, and the GET requests from S3.  The rest are relatively fixed costs - just the cost of having a reliable system with a lot of data in it.

As far as donations are concerned, at the moment, my guess is that the vast majority of users have absolutely no idea that Chumby is even gone.  I get several customer support emails a day, and all but a handful clearly think that there's still a company here.  Most users have chumbys that have been cranking along completely normally.

I think it's a tribute to the guys who wrote the servers and the ops guys that set them up.  The vast majority of outages that have happened over the last year were not the servers themselves, but problems at Amazon.  I think only one outage was actually our fault: the main analytics system was shut down at one point, but the main server was still collecting the data - without the analytics system to pull that data out and delete it, it was accumulating and eventually exceeded the allocated storage of the database.  The system was changed to throw the data away and everything went back to normal.

I have no idea how customers will react when the change happens. It may be that some significant fraction of the 40K active devices are shut down and we never see them again, which would be a shame.

Note that I am not ruling out supporting widgets ever again. It's just not the highest priority.  I have something like six weeks to get the new systems in place to keep the ecosystem from dying completely.  After that we can discuss ways to restore the widgets.

We had a design at one point where the widgets were shared peer-to-peer.  A widget publisher would put the widgets on a USB dongle in their device, with a little manifest file, and using a Napster-like system, the widgets would be made available and passed around between the devices, and stored locally.  We had figured out a way to allow users to "depublish" a widget, as well as make updates with some level of confidence that all of the peers would get the updates.  Maybe that can be resurrected it would require "Classic" owners to always have a dongle in the back - there's insufficient storage in a Classic to do this.

Maybe the solution is as has been said.   Build an easier config system for Zurk's code and a temporary 'migration' server to let users download the widgets

So you'd boot the zurk stuff.  A 'control panel' would let you ask for specific widgets, it would download them, and configure them locally.  One time, no frills setup.  You want to add/change widgets, run the migrator again from scratch.

What do you think?

Well, it's not quite that simple - many of the widgets have several different versions depending upon what device they're for - in some cases as many as four or five (3.5", 8", Android, TV, etc).  The current system takes care of managing all of the various versions.

The system currently holds about 11,000 widgets, most of which are not public.  The private widgets are most typically development versions, but some people have widgets playing on their devices that they made entirely for themselves.

We also must respect the rights of the widget authors - they must have the right to update or remove their widgets when they no longer wish to support them.  Many of the widgets hit servers - without the ability to remove or update, the author may be stuck without the ability to change or terminate their service.

There's also the issue of configuration - the custom configurations for the widgets (things like the zipcode for the weather widgets) is stored on the server - we'd need to still have a way to configure and serve this customization data.

To support all this, I would need to restore a significant fraction of the current database - which brings us back to where we started.

First of all, let me also say "Thanks" for Duane keeping things alive for the last year.

I have one suggestion for providing access for widgets without running up a large bandwidth bill, but this suggestion is more suitable for making Zurk's offline stuff work well, rather than "stock" chumbys.

The suggestion is to package the current set of public widgets with bittorrent; that way widgets can be made available without having to have a central repository.

Alternately: they could be pushed to github (or a similar service) which are happy to host open source files.  That way you can provide access to them without having to host them yourself.

I am keen to see Chumby keep going and I would like to thank Duane for his efforts. I'd contribute or subscribe to a Chumby service, if that's what was needed to keep my Chumbies on.

There's already an existing hack for the 3.5 devices to add local widgets to a channel.  It would be fairly easy to add that to the Control Panels for the 8" devices.

I agree completely! I have a bunch of widgets on two Chumbys, but I can survive with just a handful -- alarm clock, music (and maybe weather and RSS if feasible). If it all worked for a couple of hundred a month, I'd gladly sign up for a monthly or annual "subscription"! As you say, there are surely enough of the faithful among us to cover the cost.

I do think having such a thing rolled into the "final" release of the control panel would be a good idea...

Is there any way to "broadcast" an announcement to all connected Chumbies?  This might be a good idea; if not, perhaps build something into the next control panel, so you can send a message to all Chumby users @ startup to ask them visit an announcement page (which you'd presumably set up ahead of time), letting them know that the way their Chumby works will be changing in the near future (and that Android/etc will lose support), and what they can do to continue to get the most out of it.

Personally, I'd be okay with a subscription model for the New Chumby, if it can be kept cheap