Hi @PaulK
ACW’s instance of WBIA is on a physical server in our data center cabinet with access to four GPUs on the machine (though WBIA only takes advantage of one). It has a fair amount of power behind it, and WBIA runs on a different server than Wildbook entirely. That said, bulk import absolutely does a lot of work on both sides.
The next major step up in performance would be load balancing with multiple WBIA servers able to handle increased workloads, which we do plan on working on. That work has been recently funded, but it is currently scheduled.
One thing to keep in mind with the time estimate: it is a reflection of past work only and doesn’t make a prediction. The last job may have taken 18 minutes to turnaround, but if the queue is empty, it may not take 18 minutes for a new job. Currently, the time estimate is noting that the turaround time is about 15 minutes, but in fact there is nothing in the queue, so a single ID job is likely to complete much, much faster than that. I’ll look to see if we can age out our better qualify that estimate when the queue becomes empty and the estimate no longer applies.
Thanks,
Jason