0

We currently are facing problems due to high amount of cached objects. We cache Data from an ERP system (for an Online Shop) and IIS will refresh the webpage as it reaches the maximum amount of memory and we loose all cashed objects. As this makes the idea of caching a little bit problematic we are searching for a solution to cache the objects with a different solution.

I have found AppFabric from Microsoft as it is already included into our Windows Server licenses to be a pretty neat solution.

How ever I still fear that we will have enormous performance problems when using AppFabric Velocity instead of the MemoryCache Class (our current solution for Caching).

So my question is now, is this a solution for our problem or am I over-thinking here and is the performance of AppFabric fast enough?

Cybermaxs
  • 24,378
  • 8
  • 83
  • 112
Kevin Boss
  • 145
  • 12

2 Answers2

2

Grid Dynamics did a great report on using AppFabric here. While I don't know the numbers for your specific cache operations, the report showed great numbers performance wise for AppFabric. In one test, they wanted to see how the size of the cache impacted the cache operations performance. When just reading the data, it had little to no impact on the cache operations performance. When updating, there was impact on cache operations performance, but not a ridiculous amount. When testing object size and performance, obviously, larger objects lowered the performance (throughput performance here). Overall, the report has solid tests and statistics that show that the performance of AppFabric Cache is excellent.

No, Grid Dynamics does not compare the results to other products, but they do show you what the performance of AppFabric Cache is like in different tests. They have a particularly useful Appendix section that can provide details to help people in different usage scenarios.

As always, using a solution that is not on the same machine as the IIS instance will add a little bit of time to the fetching of session data from the cache, but we are talking a small amount of time.

Community
  • 1
  • 1
joncarl
  • 642
  • 1
  • 6
  • 18
0

If I am understanding your situation than there are object caching solutions available that let you cache objects in memory and expire them according to your application logic or when the cache starts filling up.

Appfabric is not a very mature product in this regard especially when talking about an "inproc" cache. You'd need a client cache, which really is a subset of the distributed cache (meaning all the cached objects) that resides "in proc" and is kept synchronized with the distributed cache.

One solution that I'd recommend is to use NCache as a distributed cache and use its clinet caching feature for your ERP objects.

Andy H
  • 132
  • 3
  • The problem is that we need to have this objecta cached at all time...so we can not throw away some when space is running out. – Kevin Boss May 22 '13 at 15:45
  • Actually, you don't have to expire the objects at all. Just have enough memory in each distributed cache server and enough cache servers to give you the desired storage space. For example, NCache has Partitioned Cache topology that lets you grow your cache size as much as you want simply by adding more cache servers. The bottom line is that you want to use a distributed cache that is rich in features and this is where NCache stands out well above all others. I'll again urge you to take a quick look at it here: http://www.alachisoft.com – Andy H May 23 '13 at 06:12
  • I will, we actually have some team members with experience in NCache so it is also an option. AppFabric was just another solution. The question was more of the performance of this caching solutions. Still, thank you :) – Kevin Boss May 23 '13 at 11:43
  • My bad! :) when you talked about MemoryCache I assumed you were talking about an "in proc" cache and that you were not sure that you would get the same performance out of AppFabric. So based on that I was pitching NCache :) because it has a client cache feature that helps avoid making trips to the main cache by keeping a subset of data from the main cache closest to the application thus bosting performance. But anyway Now I get it that you were looking for just performance numbers – Andy H May 24 '13 at 06:10