I am running a Production that has 127 tiles and one of the tiles has failed.The aerotraingulation completed on 3524 photos with 70.8 gigapixels. The machine I am running the production on is a Azure VM withWindows Server 2016, Intel Xeon 24 cores, 224 GB RAM, 4 Nvidia M60 graphics with 8 GB/card. I have optimized the VM to use the M60card as the main GPU as well as turned up the 3D settings on the card for maximum performance. I have attached the log file of the failed tile.
I started the production 2 days ago and it is on tile 21 of 127.127Tiles-20GBRAM - Tile_+005_+009.log
The aerotriangulation completed without issues and in a timely manner. The issue I started with was the amount of time it was taking during the production time to create a 3MX. The reason I created 127 tiles, I wanted to limit the amount of RAM used. With 127 tiles the RAM usage was around 100 GB. I was able to resubmit the failed tile. The project looks good but in some areas we did not have enough photos or proper overlap.
It sounds like a successful project. How long did the production end up taking? Do you think it's possible to break up a project into local aerotriangulation and Azure productions, and do you think it would be an advantage to do so? Are you using Azure to host the published 3MX in the cloud?
Looking at your project, is it right to assume 300 GB of data had to be uploaded?
The production took about 10 days. The total size of the photos was around 350 GB. The photos were a shot using 2 different cameras so that was a challenge stitching them together. What made that possible was importing tie points from actual ground shots. In the future, if we have a project that size, we will use multiple Azure machines to run the production. We will need to update our CC license so that we can leverage multiple processess.
In the future, if we have a project that size, we will use multiple Azure machines to run the production. We will need to update our CC license so that we can leverage multiple processes.
Very cool. Running multiple versions of that Azure set up on a job queue sounds awesome
Hey. Ive installed Master Center update 9 - v4.4.9.516 into Amazon AWS, disabled default monitor/adapter. It is using the right card but I cannot see power increase comparing systems 1 to 4 by performance. Running on Intel Xeon 2.3 hz 16 cores / 122 gb rams and 1 tesla m60 VM setup. Maybe there is something I do not know as a newby? because I asume it should run a lot faster.