cache raster tiles in server vs. desktop

3486
7
02-12-2014 01:07 AM
Labels (1)
larryzhang1
New Contributor III
Team,

Considering the current mechanism of caching images with ArcGIS Server not manageable very well in our enterprise environment (take too long time to cache, tight security, NAS storage over the network enabled with anti-virus scanning, etc.), we realize that more reliable cache approach is: ???To cache /update raster tiles with MD directly in desktop???.

The question is:

What are major differences between the server cache tiles (with MXD or MD) and desktop cache tiles (with MD), except for more flexible capability in the server (saying, cache individual level(s) on defined AOI)?

In other word, is there any similarity on cache algorithms between both?

To me, it looks that the compression algorithm on both is different (even though we use the same), because the size of the final cache tiles is different on the same MD??? Is it true? or something missing?

Pls advise...

Best regards,
0 Kudos
7 Replies
PeterBecker
Esri Regular Contributor
The cache created by server and desktop are the same. Both create by default Compact Cache that bundles the large number of JPG and PNG tiles into a smaller number of bundles. The data volume is also the same and is dependant on the levels, extent and the compresion quality (when using JPEG or Mixed).
Note that even on desktop there are two differnent ways of generating the cache, either caching a Map or caching an image. If the content is image, then using the method of caching the imagery (as a single raster, raster layer or Mosaic Dataset) is preferable as it generates the cache from the highest level (smallest pixel size) first. This means that the system can generate overview by subsamling the higher resolution imagery. There was a recent Live Training Seminar on caching imagery.
0 Kudos
larryzhang1
New Contributor III
...

There was a recent Live Training Seminar on caching imagery.


Thx for video, Peter,

Yes, both of the final cache size is same/close, when the schema /GP is 'properly' used.

With the cache mechanism under desktop, it takes much less time to do tiling /updating, which makes cache operation much easily manageable in enterprise, no matter where massive raster sources locate at, SAN or NAS (UNC) �?�


Regards,


++++

By the way, while performing the desktop caching (onto MD),  it looks that the schema should be directly generated from MD, rather than MXD (or arcgisonline), on hands-on comparisons. Otherwise, it (desktop) produces differrnt sizes of the final compression caches (attachment)...


[INDENT] 10%~20% difference on same demo MD, when the schema used differently[/INDENT]


Also, different GP tools produce variant cache sizes, even using the same schema from MD, saying, GP like �??Manage Tile Cache�?? vs. �??Generate Tile Cache�?? (attachment 3)


[INDENT] The size of caches 15 GB vs. 2 GB on same above MD, when using different GP...[/INDENT]



Not sure why ...

(It should be treated as an issue, if true).
0 Kudos
larryzhang1
New Contributor III
Hello, Peter,

Is there a way to configure image service displaying �??Dynamically from the data�?? or �??Using tiles from a cache folder�??, in particular, in client-side as an option?

The background /the scenario are:


[INDENT]When the image service is published with �??Using tiles from a cache�?? in the server (for performance), we also want the same image service fully functional with �??Dynamically from the data�?? (only for analysis,  as an option in desktop, when needed).

As recalled, someone from your team showed us a way to configure that. Not sure if correct�?� [/INDENT]


Otherwise, we have to re-publish as separate image service with different name for this purpose, which leads to duplication�?�

Regards,
0 Kudos
PeterBecker
Esri Regular Contributor
Yes. If you publish an image service then you will also see a 'Caching' tab. This enables an image service to be directly cached by the server. This is very similar to caching a feature service in that a cache of the imagery is also created. The cache can be fully created or in the advanced option there is a 'Create tiles on demand' option that causes the system to generate cached tiles only when a user first goes to an area. If an image service with a cache is added to Desktop then in the layer properties one can select if to access the cache of the dynamic image service. In web applications the app needs to define what mode to request the service in. Currently in a webmap you can not switch between the two modes.
0 Kudos
larryzhang1
New Contributor III
Yes.

...

If an image service with a cache is added to Desktop then in the layer properties one can select if to access the cache of the dynamic image service. In web applications the app needs to define what mode to request the service in. Currently in a webmap you can not switch between the two modes.



Thx, Peter,

It works.

The view mode in desktop 10.1/10.2 can effectively switch from the cache mode to regular image service, as expected...

Great job, thx a lot, team!
0 Kudos
JoshWhite
Regular Contributor III
What is MD?
0 Kudos
larryzhang1
New Contributor III
What is MD?



Yes, it looks lazy.

MD is here referred to Mosaic Dataset, just like PM (Private Message) and GEE (Google Earth Enterprise) ...
0 Kudos