Where should I put my mosaic dataset

[ModyBuchbinder]

Hello all

I have a file Geodatabase with a big mosaic (over 500k tiles) that I need to publish as a map/image service.

My Server is a virtual machine.

All raster data reside on a SAN/NAS disks

The FGDB itself is not too big so I can put it in any where I want.

Where will be the best place?

1. On the local C: drive (even it is virtual)
2. On the local C: of the machine that runs the virtual machine
3. On the SAN/NAS drive (which one is better)?
4. Anywhere else (SSD drive)?

Thanks

Mody

[LanceCole]

We place our mosaic image GDBs on a share and have the users copy the file to their local drive.  This provides multi user access and a slight performance gain.  One caveat is the users either have to use the same drive mapped to where the actual images are shared (ie the "I" drive) that the mosaic was created from.  Another option is to create the mosaic data set using a ​UNC network path and provide user read access.  The original images and overview files are stored on our systems SAN or NAS.

[JonathanQuinn]

If you're concerned with performance, reducing the amount of requests that need to go over the network is going to be key.  You'll want to put the file geodatabase on the local Server machine so that at least to reach the mosaic dataset, ArcGIS Server doesn't need to reach the network.  It will have to to retrieve the actual tiles through the network, but that seems unavoidable given the amount of data you're serving.  Following Lance's suggestion, you can still set up a UNC path for the data so other users can use it in their Desktop applications.

[ModyBuchbinder]

Thanks for your answers.

The two things I cannot change is that the mosaic is served as a service and the server is virtual.

The C: drive is not the same as the C for physical machine.

This is the reason for the question.

Thanks

[JonathanQuinn]

Right, it's understandable that the C:\ drive for a virtual machine is not truly on the local file system, but for the communication between the Server and the mosaic dataset, it'll be quicker than if it were set to a UNC path.