3

I am writing a messaging application for Android. Because 30% of Android users are still on 2.1, I have decided we will not use Google's C2DM system for push notifications, which requires 2.2+.

Does anyone have experience building medium to large scale services on android that require push notifications? I'm interested in things such as:

  • How many users can reliably be supported per server or per IP address?
  • How difficult is it to increase capacity as you go?
  • How long will it take to set the server up?
  • How reliably does it deliver messages?
  • How quickly does it deliver messages?

I know there are a few popular solutions out there such as MQTT, Deacon, Xtify and Urban Airship, but I cannot seem to find reliable data about the above topics from people who have actually implemented these solutions in capacity.

plowman
  • 13,335
  • 8
  • 53
  • 53

2 Answers2

2

Urban Airship's blog has a couple of entries that describe how they are scaling their Android push solution for pre-2.2 devices:

gnuf
  • 2,722
  • 1
  • 25
  • 32
1

Think in the future, I highly recomend you take a look at the time it has taken for Android OS' to evolve on phones and study that against the cost of using other non-official solutions, mainly, the question you should ask yourself is, how long will 2.1 phones be on the market vs how much will it cost to run other solutions.

Do you want to break your easy solution for an "exception" to the rule? It really depends on what you want, I see people still developing for 1.6 and such, when IMO there is really such a small market share to attend there that its not worth breaking the easier and more useful 2.1 API.

As for designing your own push server, you are talking about a project that will probably take longer than 2.1's useful life. I mean from scratch.

I cant give you any insight on the solutions you have mentioned though.

blindstuff
  • 18,298
  • 10
  • 47
  • 48
  • 2
    The problem is that at its current rate of decline, it will still be another 6 months before 2.1 will be negligible in the US, and another 9-12 before it will be so in the developing world. I have no problem switching to C2DM at a later date, but think it's important to have a working 2.1 solution for now. As for designing my own server from scratch, I will definitely not be doing that. Solutions like MQTT are already written and will more or less work out of the box. – plowman Mar 09 '11 at 19:04