I am using EhCache framework to cache application data and thinking to use JGroups cache replication to replicate cache in a clustered environment.
Is it really an industry standard for cache replication in clustered environment? Or, there can be other better options that I should think about. Please notice that I am not using any centralized cache server at this point of time. I have already done POC on JGroups Cache Replication. Could you please share your experience in terms of its robustness and major concerns? What are the pros cons of using JGroups for cache replication?
I am using Jgroups for clustering various application nodes. We have our own cache implementation which uses the underlying Jgroups for data replication/distribution. So far it is working fine without issues.
Could you please check 'infinispan'. It's a distributed cache which uses Jgroups for handling cluster.
Related
We plan to use Ehcache in our application that resides in a cluster. However we do not want to use Terracotta SA, what are the other options for cache replication across all the servers in cluster?
The quick answer is: No.
With Ehcache 2 you could, but it wasn't really reliable. With Ehcache 3, the clustered is done using a terracotta server. This has always been the most reliable way.
Note that the terracotta server is open source.
I am trying to deploy a spring-security server, with Redis as token store.
In order to have some redundancy in Redis, we want to deploy it as a cluster.
The problem is Jedis, which is used by spring security as underlying library, doesn't support pipelining in cluster mode, but spring security uses pipelining.
My question is how can I solve this situation. More precisely:
1- Should I use another mode of deployment form Redis? What actually works.
2- Can I somehow force spring security to use reddison for connecting to Resid?
Please adivse.
If you want redundancy, use replication (master/slave) not cluster.
If you have more data than RAM on a machine, use cluster.
If you have more data than RAM on a machine and want redundancy, use cluster with replication.
Jedis supports replication with sentinel, so give that a go unless you have a lot of data. Some more info on usage here: https://github.com/xetorthio/jedis/issues/725
What's the difference between infinispan local cache and distributed cache. I know that in local cache no cluster can be formed whereas in distributed cache we can form the cluster. My doubt is whether the distributed cache data will store in our running application heap or external to the application. Also, whether it is possible to obtain the data stored in infinispan running in one jvm by application running in another jvm.
Infinispan stores the data in the application heap. It also can persist the data if you configure a CacheStore (for example a database). You have the details here: Infinispan Persistence
About the second question: yes, Infinispan know where the data is stored and fetches from other nodes (or jvms) if no copy is available local. If you need more details, you can check the documentation: Infinispan Clustering
I took an old web application recently. The App will save a lot of object into Ehcache at init time.And then obtain the object from the cache. And now, the object has increased many times. Ehcache in the app can not meet the requirement. So, we consider using a distributed cache. We will set some strategies and let the objects save in different cache server.
Redis and Hazelcast are both good. Question is, the Redis and Hazelcast compared to the previous Ehcache, Redis and Hazelcast must Serialize objects. May be it will consume more time.
So which is better?
Or are there no other better alternatives?
Thanks in advance.
What are your requirements? Hazelcast has pluggable serialization and there are serialization libraries like FlatBuffers that are exceedingly fast if needed.
Hazelcast is ideal if you are Java centric, of course you can also access Hazelcast via .Net CLR languages if you are using Hazelcast Enterprise.
Hazelcast is more than just a simple distributed cache, you can also leverage the distributed CPUs in your cluster and execute processes directly in the grid, which can result in very high performance.
If you are multi-language and dont need to execute code on top of distributed data (in-memory executor service, mapreduce or entryprocessors) Redis would make sense.
Have a look at Redisson - Redis based In-Memory Data Grid is a good alternative for Ehcache. Provides a lot data serialization codecs like:
Jackson JSON, Avro, Smile, CBOR, MsgPack, Amazon Ion, Kryo, JDK Serialization, LZ4 and Snappy.
It also allows to store Map data in external storage along with Redis store using Read-through, Write-through and Write-behind strategies.
Offers distributed MapReduce and Executor/Scheduler services.
I am using jboss-cache for session replication in jboss 4.0.4. How do I use a different distributed cache(Terracotta or any other) in place of jboss cache?
You may need to repackage jars and configuration in your jboss server, but Tomcat has its own clustering and session replication support. Not that I recommend this approach. Any specific reason you don't want to use the built in session replication ? You may also find improvements in clustering and session replication in later versions of JBoss 4.x.
Here is documentation for Terracotta For Web Sessions. If you disable clustering in your JBoss server and implement this, you should be good to go.