Untitled Document
Page 1 of 2 12 LastLast
Results 1 to 10 of 15

Thread: ShipRush UPS Pro v12 - SOAP performance issue

  1. #1
    We have a client that has recently updated to v12 (from v10 we believe) and is experiencing significantly degraded performance.

    This client has been using ShipRush Pro for UPS (via our app and the SOAP method) for many years and, due to their high seasonal volume, have made every effort to fully optimize its performance. These efforts include running regular database maintenance (performance degrades rapidly after ~1000 shipments), a dedicated high power machine for the ShipRush server where ShipRush runs on an SSD, ample memory, etc. In their busy season they go on to use multiple servers and ShipRush instances with a load balancer (currently off.)

    In short, they have significant experience with optimizing ShipRush performance since it has such a huge impact on their order throughput which can exceed 10,000 packages a day during holiday time. Their past performance produces SOAP responses (without address validation or printing, just the SOAP response) in the 1-2 second range. Most are air shipments.

    With the recent and necessitated move to v12, they have been evaluating performance and have reported significant declines. We're working with them today to learn what we can and make sure there is nothing obvious amiss. Before we get too far chasing this, I wanted to run it by you and see if this is known, expected, or can be mitigated. Should they still expect to be able to obtain 1-2 second response times on an optimized server?

    Our understanding in the past was that performance was strongly impacted by disk IO, hence they are running SSD. In v12 has this perhaps been moved to a web service which could be hampering performance?

    Thanks,

    Russ Horton
    Product Manager
    NewHaven Software

  2. #2

    Join Date
    Apr 2004
    Posts
    16,500
    SYSOPS is offline
    Asking development to review further. Case 68171. We will post with an update to this question when available.

    Thank you,
    SYSOPS
    Have Questions? Ask the ShipRush Knowledge Base: https://shiprush.com/kbase
    Reply With Quote

  3. #3

    Join Date
    Apr 2004
    Posts
    16,500
    SYSOPS is offline
    From development:

    There is no new web service IO in v12.

    We expect performance to be the same as v10

    NOTE that you can run v10 and v12 in parallel, on diff VMs, either discretely or behind a load balancer.

    All that said, we expect that v12 can give the same performance. Checking the shiprushserver.log for errors is a good first step.

    SYSOPS
    Have Questions? Ask the ShipRush Knowledge Base: https://shiprush.com/kbase
    Reply With Quote

  4. #4
    It turns out that their issue was misrepresented to us. Digging further, their problem is not with the upgrade from v10 to v12.

    Their issue is that they are suffering the rapid database performance degredation when processing large volumes of shipments (1000+). This has been the case since they started on ShipRush (v6?) and to date they have tried to make the best of it with hardware, process, and load balancing. In prep for the upcoming holiday season they recently bought a new beefy server (dedicated to ShipRush) to increase performance further but it did not help (which is what prompted their contacting us.)

    This client has processes in place already to run the database maintenance with regularity but their volume continues to grow and they need to be able to process larger batches of labels without running database maintence.

    We haven't broached this subject with you in many years. Are there any optimizations or new understandings of this performance issue that we can explore with them?

  5. #5

    Join Date
    Apr 2004
    Posts
    16,500
    SYSOPS is offline
    The latest discussion about optimizing throughput is spelled out here:

    http://forums.zfirm.com/showthread.p...the-throughput!

    It sounds as though you're already doing those things. I am not aware of any new understandings to this question.

    SYSOPS
    Have Questions? Ask the ShipRush Knowledge Base: https://shiprush.com/kbase
    Reply With Quote

  6. #6
    It seems I don't have access to the linked post, not sure why.

    I also cannot get the following page to load, something I found by search - https://shiprush.com/assets/images/S...d-Balancer.png

    You're right though, they have been using a load balancer already and probably nothing new there since the load balancer solution was first posted. Correct me if I'm wrong.

    I suppose the core issue to revisit is what you see as the most impactful issue with SOAP server performance. Our understanding was disk read time and that mem and processor had little impact beyond satisfying the needs of the OS but let me know what you think.

    We're trying to work with them to compare performance between the two dedicated ShipRush servers they have and how to benchmark to find the best one or if alternate hardware should be explored.

  7. #7

    Join Date
    Apr 2004
    Posts
    16,500
    SYSOPS is offline
    Please try those links again. They are in the SDK area, and are closed off by default.

    SYSOPS
    Have Questions? Ask the ShipRush Knowledge Base: https://shiprush.com/kbase
    Reply With Quote

  8. #8
    Thanks, that did the trick. It does help establish that they are not achieving optimal performance even after database maintenance. They are currently reporting a best case processing time around four seconds a label (single server). My recollection was they had achieved 2 seconds in the past and your post says you were getting 3 seconds (20 per minute.) We're trying to get them to focus on ensuring optimal performance on small batches (with database maintenance run) before we start worrying about processing at scale (which I think will just be additional servers in the load balancer.)

    To that end, can you confirm what you think are the biggest contributing factors for SOAP performance? Is there more to it than disk read time?

  9. #9

    Join Date
    Apr 2004
    Location
    Seattle
    Posts
    8,959
    Blog Entries
    1
    SYSOPR is offline
    The performance topic has a lot of aspects to it.

    The only way to really engage it is with a test/reference system.

    There are a few ways to slice it. One way would be to have a VM with the older SR in it, and the test tool, review throughput, and then a snapshot, and then same VM with newer SR in it.

    The only way to really engage the question is with a test system.

    (We are assuming that the run time environment [SR db size, system performance, etc] are the same or better.)
    -----------------------------------
    The Blog: http://forums.zfirm.com/blog.php

    YouTube channel: http://youtube.com/shiprush

    Thank you for using ShipRush!
    Reply With Quote

  10. #10

    Join Date
    Apr 2004
    Location
    Seattle
    Posts
    8,959
    Blog Entries
    1
    SYSOPR is offline
    ALSO, I should add, that you can route the SOAP payload directly to the ShipRush cloud API. That can give you great parallelism.

    This may be worth considering.
    -----------------------------------
    The Blog: http://forums.zfirm.com/blog.php

    YouTube channel: http://youtube.com/shiprush

    Thank you for using ShipRush!
    Reply With Quote

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •