Ultra pipeline response time

Hi everyone,

I’m testing the Ultra feature performance and I’m finding a results that are not so satisfying.

  1. Using the new release 4.22 and even the old release 4.21 (including the patches):
    Simple pipeline containing only JSON Generator snap with the default Hello World output, Bearer token authentication, no query parameters, no additional URL paths in the request. I’m receiving minimum response time of 900 ms (it is sometimes going up to few seconds).
    I have tested the same pipeline on few organizations and using two REST clients (Postman, Advanced REST Client) but the results are same.

  2. Using the older release version 4.20. Few month ago, (I believe it was March) I was testing the same pipeline configuration in Ultra mode and I was receiving minimum response time of 400 ms.

I don’t believe that response time of 900 ms is acceptable for such a simple process.

Are there any tips and tricks that should be applied on the nodes/snaplex to improve the performance of the pipelines running in Ultra?
SnapLogic development team, Is this something that is due to some wrong configurations/setup on my environments or there is a change between the releases that is making an impact on the Ultra performance ?

1 Like

We also saw this performance hit in the 4.21 upgrade but mostly using PATH parameters. First call you would see almost an 8 second hit and then any subsequent calls seemed normal. Change the path param value the 8 second hit came back. We reported this to support and it was acknowledge as a bug once we proved to them it was environmental. We just tested this with 4.22 after upgrading the plex environment in our dev environment. The initial problem does seem to be addressed but we are seeing more latency than before. Since it’s dev it difficult to truly tell if latency or normal. We typically have 200ms to 800ms response times depending on our pipeline. 800ms for just a JSON generator seems excessive. Have you reported that to Support?

Hi @murphy-brian,

Thank you for your reply. I’m still struggling with the above but the problem is also addressed to SnapLogic so waiting on some update.

When you look at the execution statistics in the pipeline runtime -> Extra details, what are the runtimes shown there?
Are you using multiple instances of the Ultra pipeline?
Are you using a load balancer in front of your feedmaster ?
Is this running on a Groundplex or a Cloudplex?

Hi @cstewart

Thank you for your interest on this topic. Below are the answers of your questions.

  1. Extra details in the execution statistics

  2. I’m using only one instance for this test pipeline but it is not something that is heavy used. I just occasionally triggering the pipeline from Postman to test the performances.

  3. Yes, I can see in the Manager that the field ultra load balancer is populated with some address.

  4. This is running on Cloudplex.

Please let me know if you need more details or if you find something that should be updated in order to improve the performance.

On your feedmaster node, what do the averages look like for the task? (Dashboard, FeedMaster Node->Additional Details)
I’d suggest you make a few requests, as with any process if it gets idle for any length of time, it gets swapped out. So the first invocation may be slow, but the subsequent requests should be fast(er). If that too is in the ms range, then the loadbalancer might need to checked by the SnapLogic Infrastructure Management team. (as it is a SnapLogic managed resource)

Hi @cstewart,

Thank you for your response. Here is what I found in the Additional details section of the feed master node.

The below image is taken from the Dashboard -> Task, for the appropriate ultra task.

Yes, the first execution is always bit slower which is understandable and acceptable. The other executions are faster then the first one, but a lot slower than what we expect for such a simple pipeline. Please let me know if this should be checked by the Infrastructure team and how can I get in contact from the Infrastructure team.

Bearing in mind the roundtrip from your source (postman locally on your laptop?) to a Snaplex running in the US-West region, if you are seeing sub-second response, I’m not sure I see an issue. What kind of response do you see if you run a pipeline locally on the same Snaplex making a REST GET to your Ultra task?
If there is an issue with the load-balancer, then you should contact your SnapLogic support team.