How to save Live and Historical Vehicle Location data in Amazon DynamoDB

0
votes

I am trying to process and store location information of about 30k vehicles in DynamoDB.

  • The Live information will be stored every 10 seconds for each vehicle.
  • It will be processed to Live track single vehicle
  • The info will also be used to get nearby vehicles given a point and radius.
  • I will use python code

I am following following strategy:

Hist table keeps device historical information

  • Data will be: deviceid (partition key), lat, lon, timestamp (sortkey), geohash

Live table keeps live info

  • Data will be: deviceid (partition key), lat, lon, timestamp, geohash(sortkey)

  • geo hash helps in searching nearby vehicles

  • the table information will be updated such that only 30k data will present, one for each vehicle at a given time.

How can I improve on scaling of the Hist and Live table read and write given write is done every 10 seconds?

1
pythonamazon-web-servicesamazon-dynamodbgeohashing
What scaling problems are you current experiencing? – John Rotenstein

1 Answers

1
votes

You can specify the Read and Write throughput of DynamoDB tables upon creation and you can modify throughput later if necessary. In addition, DynamoDB allows bursting beyond these defined throughput limits.

To obtain the full throughput of your tables, use a wide range of Partition Keys so that requests are distributed to many different servers.

DynamoDB now also supports Auto Scaling, so it can automatically scale based upon usage.

For improved scaling of eventually-consistent Reads, you can also use In-Memory Acceleration with DAX.

In situations of bursty writes (where there might be insufficient Write throughput), some AWS users use an Amazon SQS queue to temporarily store data after received a Throttling error, with a backend process that later reads these messages and inserts them into DynamoDB. This allows tables to be provisioned for average throughput rather than peak throughput.

Bottom line: You should be able to avoid most scaling issues by increasing the throughput of your tables. Other techniques (such as those detailed above) can provide even more scale.