-2

Possible Duplicate:
Can you help me with my capacity planning?

Having worked for weeks on a new mobile app product, we are finally preparing a public launch in the coming weeks. However we wonder if our Europe based servers will be good enough (latency, etc.) for american users. Tested on dotcom-monitor.com and we got around 150ms for CA. Is it good enough?

We are exploring EC2 options but it is way more costly than local dedicated hosting, and I've been told that you should not care / work that much on scalability at early stages.

Considering the time&cost it would take to go to a replicated EC2 worldwide setup, we are not sure wether we want to commit on this too soon.


  • More info about the BW usage :

On startup, the app performs 3 ajax request to our API (1 login + info & 2 indexes). Typical bw usage would be around 5k (gziped). From europe the biggest request takes 62ms connecting (34ms ssl) & 140ms processing. After that, the app is 100% usable, it will only perform one request at a time on user interaction / push.

Olivier
  • 415
  • 3
  • 5
  • 14
  • 3
    Do you have documentation for your internal use that outlines what is and what is not acceptable performance for this app? I'd suggest that rather than looking at numbers, you need a real-world test of how well your app performs under those conditions. – Rob Moir Jul 16 '12 at 07:36
  • 4
    How could anybody here know an answer to your question? We don't know your product, we don't know your requirements. – Oliver Jul 16 '12 at 07:38
  • Just edited to be more specific about the BW usage. – Olivier Jul 16 '12 at 07:44
  • Besides what has already been said, mobile users tend to have much higher tolerances for latency due to 3G (waiting until the train moves to a better area, low signal, tunnels, etc) and less CPU / RAM available in their devices. Personally, I usually wait 10 - 20 seconds before trying an action again when I'm commuting. – Ladadadada Jul 16 '12 at 08:09

2 Answers2

3

I'm not sure what you app does, but if it can be proxied the charles might be able to simulator latency and see if you application 'feels' usable with varying amount of 'lag'.

http://www.charlesproxy.com/documentation/proxying/throttling/

If that's not suitable, other network test tools are available.

The Unix Janitor
  • 2,388
  • 14
  • 13
2
PS> ping stackoverflow.com

Pinging stackoverflow.com [64.34.119.12] with 32 bytes of data:
Reply from 64.34.119.12: bytes=32 time=221ms TTL=56
Reply from 64.34.119.12: bytes=32 time=227ms TTL=56
Reply from 64.34.119.12: bytes=32 time=223ms TTL=56
Reply from 64.34.119.12: bytes=32 time=227ms TTL=56

Ping statistics for 64.34.119.12:
    Packets: Sent = 4, Received = 4, Lost = 0 (0% loss),
Approximate round trip times in milli-seconds:
    Minimum = 221ms, Maximum = 227ms, Average = 224ms

150ms is normal for many services on the internet when going across an ocean.

Unless your application is inherently particularly sensitive to latency, 150ms shouldn't be a problem.

(Note, there are plenty of parts of the world where 500ms would be a typical latency.)

Richard
  • 5,309
  • 1
  • 22
  • 20
  • Wow, thats some sick latencies you are talking about. Personally, when it gets over 100ms I begin to worry. Although 150 shouldnt be that much of an issue - but it also depends on the requesters internet connection. – Frederik Jul 16 '12 at 07:46
  • @FrederikNielsen Not currently helped by doing a download (so looks bad currently). But the point is really in the final two paragraphs (I suspect the OP is based in North America, and assumes what is true in the US in the norm elsewhere.) – Richard Jul 16 '12 at 07:49