Caching time / Not using max number of instances

JulieKusic · ‎09-28-2012

I am using 10.1. How can I easily tell how long a cache runs? In 9.3.1, when the cache completed, the message box would indicate how long it ran.

Also, I am using a 16 core machine dedicated to caching only. I set the max instances to 17 but it only ever uses 1. Why??

RoyceSimpson · ‎09-28-2012

I am using 10.1. How can I easily tell how long a cache runs? In 9.3.1, when the cache completed, the message box would indicate how long it ran.

Also, I am using a 16 core machine dedicated to caching only. I set the max instances to 17 but it only ever uses 1. Why??

I'm having similar issues. I've got multiple machines and a couple "clusters". One cluster has only one machine in it and the other cluster has a bunch. I'm trying to use that second cluster for caching. When I setup a map service and tell it to live in the second cluster, then run the map cache... it doesn't use any of those cpus (even though the map service is reporting the correct number of "max instances") and uses only cpu's from the machine in the first cluster. Very odd. I've checked and re-checked that the service is setup properly and is being hosted by the second cluster. The "instances running" counts in both the web manager and ArcCatalog are very confusing as well and they state conflicting and erroneous map service instance counts when the cache is running.

Good luck on this. I've posted a similar thread on this forum as well.
-r

RoyceSimpson · ‎09-28-2012

Ok, I'm up and running at full caching capacity.

You need to make sure your CachingTools services is pointed at the correct cluster. I had mine pointed at my wrong one, which only has one machine in it. I pointed it over to my "caching cluster" and now my map service, which is hosted on the first single machine cluster, is caching at full multi-machine cpu capacity. Pretty cool. Really cool actually.
-r

JulieKusic · ‎09-28-2012

Ok, I'm up and running at full caching capacity.

You need to make sure your CachingTools services is pointed at the correct cluster. I had mine pointed at my wrong one, which only has one machine in it. I pointed it over to my "caching cluster" and now my map service, which is hosted on the first single machine cluster, is caching at full multi-machine cpu capacity. Pretty cool. Really cool actually.
-r

I only am using one machine, which has 16 cores, and I'm using the "default" cluster under the Caching Tools. Glad you are up and running! I'm going to poke around a little more

RoyceSimpson · ‎09-28-2012

What is the "max instances" value for your cachingtools service? If it's 16... and you set your cache to process with 16... and you are only getting one cpu to process... yeah, that is odd.

JulieKusic · ‎09-28-2012

What is the "max instances" value for your cachingtools service? If it's 16... and you set your cache to process with 16... and you are only getting one cpu to process... yeah, that is odd.

My max instances in the Caching Tools Service is 17. My min was 0, so I changed that to 1... not sure if that will do anything.. I'm going to try to run again

RoyceSimpson · ‎09-28-2012

My max instances in the Caching Tools Service is 17. My min was 0, so I changed that to 1... not sure if that will do anything.. I'm going to try to run again

I'd drop back to 15 or 16... one more than your machine has cores was the 10.0 model. Seems like the new way is "one less". Actually... it's "one less" than the total... so if you are running 4x4core machines (sixteen total) you'd cache with 15 instances. When you have a single 16 core machine.. I'm not sure what you'd set the caching instances to.

By the way, you don't have to set the max instances per machine for the map service itself to the number of instances you want to cache at... only the "cachingtools" service needs to be set to that high number.

See this ags 10.1 help doc for the deats.
http://resources.arcgis.com/en/help/main/10.1/#/Allocation_of_server_resources_to_caching/0154000003...

JulieKusic · ‎09-28-2012

I'd drop back to 15 or 16... one more than your machine has cores was the 10.0 model. Seems like the new way is "one less". Actually... it's "one less" than the total... so if you are running 4x4core machines (sixteen total) you'd cache with 15 instances. When you have a single 16 core machine.. I'm not sure what you'd set the caching instances to.

By the way, you don't have to set the max instances per machine for the map service itself to the number of instances you want to cache at... only the "cachingtools" service needs to be set to that high number.

Ok I'll try that. Thanks for your input!!

RoyceSimpson · ‎10-01-2012

I figured I'd piggie back on this thread since the title has my issue in it.

I've got 39 instances available for caching (I have 40 cpu's dedicated: 39 for the caching + 1 for the controller instance) a service and yet when I run the cache the number of instances in use seems to fluctuate between 15 and 35 but never gets to 39 and usually hovers somewhere in the 20's.

On the face of it, it seems like the caching job isn't running at full throttle. What can be done to get the cache job to run as close to the 39 instances for as long a sustained period of time as possible?

MegPeterson · ‎10-05-2012

I am using 10.1. How can I easily tell how long a cache runs? In 9.3.1, when the cache completed, the message box would indicate how long it ran.

Also, I am using a 16 core machine dedicated to caching only. I set the max instances to 17 but it only ever uses 1. Why??

Hi Julie - You can see how long the caches take by looking at the Geoprocessing menu/Results in ArcCatalog and expand down the Messages information. I generally run my cache scales manually - 1 at a time - so I can see how long each scale is expected to take.

Hope that helps, Meg