Requests slower than expected and request handling sequence

Rakesh.Singh

A couple of questions on behalf of DSS prospect

1. Our
developers have been doing some performance testing of DSS REST API TimeSeries
requests, this has prompted a few questions that you may be able to help
with...

We're
finding this type of requests slower than expected.

• When running a testing using
  all of the securities within the Henderson 2 year test data (where RIC
  available is about 11000 instruments), requesting data for a period of 7
  days, took 20 minutes to return everything.
• A second test using test code
  from from Thomson Reuters development community forum produces this
  result:

This
prompts our first question: Is there any throttling or other performance
limits for our test account that might explain the slower than expected speed
or is this typical? Is there anything we might try that could speed this up?

Secondly,
when reviewing the documentation we found the following:

Could
you help clarify this for us - does this mean, that requests of the same
type will be processed synchronously on the server even if a request is
asynchronous?

The
follow on from both of these, is a more general question regarding requests: what's
the most efficient way to load timeseries data for a big list of IDs and a few
years date range?

1.png

2.png

Christiaan Meihsl

Rakesh.Singh

To your first query:

Performance depends on the server load, which varies during the day, making it very difficult to give numbers. That said, those you mention do not seem unreasonable.

Note: once the data has been delivered, the next get on the location URL will deliver a 404; that is normal, because the data was delivered in the previous call.

To your second query, on:

Multiple requests for the same data type, like tick data, will be placed in one queue, whereas requests for tick data and Intraday bars use independent queues.

This means that requests of the same type are all placed in the same queue. If that queue has a lot of pending requests, processing might take more time than a request for a different type of data that will be placed in a different queue (which could have very few pending requests and therefore process faster).

For that reason, if you have to submit requests for several data types, instead of submitting one, waiting for the results, and then submitting the other, it is worth while submitting both, and checking for results as they become available.

To your general question:

For efficiency, it is best to:

• Avoid (if possible) placing requests at high server load times, like the hours just after big markets close). Weekends are best.
  
• Make few request with many instruments (within the allowed limits), instead of multiple small requests.

I'm trying to understand your use case. Will you need to make such large requests on a regular basis, maybe even daily, or is this a once-of to fill a database ?