Traffic and Bandwidh calculation

Proper forum for all questions related to xServer INTERNET = the Azure based cloud solution. This forum deals with the architecture topics such as load balancing, available profiles and standard server settings.
Please be aware that questions about core functionality such as routing, mapping, geocoding in general should be placed in the xServers dedicated forum ;-)
Post Reply
Lars.Sert
Posts: 1
Joined: Fri Feb 05, 2016 9:43 am

Traffic and Bandwidh calculation

Post by Lars.Sert »

Traffic and Bandwidh calculation

How much network traffic generates the connection from the client to PTV X Server INTERNET?

That means:

To calculate our needed Internet Bandwidth in our planned Data Center (Hosting Contract), we need to know
how much traffic the quantity of X queries at the same time generates.
For this purpose we have to calculate the average traffic per query.

For example:

• We have 50 Clients.
• Every Client is going to send max. 1 Query per minute.
• So we need 50 times the theoretical calculated amount of traffic per Query.

Would be very helpful to know these bench mark.

Many Thanks
Lars Sert
User avatar
Bernd Welter
Site Admin
Posts: 2564
Joined: Mon Apr 14, 2014 10:28 am
Contact:

Re: Traffic and Bandwidh calculation

Post by Bernd Welter »

Hello Lars,

yeah - that is definetly a challenging question. I already placed it at several departments (internal cloud providers, development, product management).

In fact we do not have benchmarks that contain information about required bandwith - obviously our internal testing suite is based on a sufficient infrastructure. From a technical point of view the required XML or JSON documents sizes vary a lot depending on various factors. I'll comment them service by service - but the conclusion is more or less "you need to test it based on your own functional requirements":
  • xlocate.findAddress: a simple geocoding call about 2KB, the response with about 10 hits produces about 20KB - which is ridiculous.
  • xlocate.findLocation: s a simple reverse geocoding also about 2KB request, 20KB response (I assume that you limit the result hit number)
  • xRoute.calculateExtendedRoute: response size depends on geographic size of the track, requested response elements and response geometry style (WKB prefered). Example: input size <3KB. International track (Luxemburg to Warscaw, 1300km) incl. polygon and texts produces about 2MB response (uncompressed). Small track (<50km) is less than 100KB response.
  • xMap.renderMap: pure basesemap request (256x256, PNG) less than 2KB (bigger if you request additional POI info in the map). Response size is about 30KB. In a tiled environment you may require a matrix of tiles for a large fullscreen experience, e.g. 6x8 tiles = about 50(!) of them. Depending on the caching effects this is like requesting 50x30KB = 1.5MB
  • xTour.planBasicTours: hard to determine - some requests deal with 1 vehicle and less than 10 orders while others try to dispatch 2000 orders over 100 vehicles... can't give a concrete value. But: with the larger scenarios the bottleneck is no longer the network bandwith but the the computation time in the backend. SO this is when it comes to asynchronuous protocols.
These tests are furthermore based on proper parametrizing paradigmas
  • do not request what you don't need (xroute.ResultListOptions, proper method)
  • use WKB format whenever it is possible
  • use JSON if possible (my tests were based on SOAP)
I hope this answer is sufficient - of course you have to multiply such values with the workload generated by your users.

Best regards
Bernd
Post Reply