I have a service based architecture where a web farm full of asp clients hit application server farm of WCF services. Obviously all the database access is done by the WCF services. Now I would like to cache my frequently used database retrieved objects using Velocity at the service tier level. I am considering to make each physical application server also part of the cache cluster.
According to Velocity documentation, if I use regions, objects are stored only at a single host. I actually wouldn't have any problem if each host kept it's own cache provided that I could somehow synchronize them.
So my questions are
If I create one region on one host is it also created on another one?
When I clear a cache region, is it cleared on one host only?
If I subscribe to a region level notification on all the hosts, can I catch events of one host on another one?
In this scenario should I use regions at all or stay away from them?
I hope my questions are clear. Actually I am more interested in a solution to my problem than answers to my questions
Yes you are right in reading the doc that the region will exists only in one host.
" I actually wouldn't have any problem if each host kept it's own cache provided that I could somehow synchronize them."
When you say synchronize, you mean when HA in enabled ? Velocity would actually take care of that if thats what you meant.
For the questions:
1. No.
2. Yes
3. Notifications will be sent to the client. So i am not sure if there is anyway to send notifications to other host.
4. Regions gives Search capabilities and takes away HA from you. In your case, you could use the advantages of HA.
Having regions not necessarily means that you don't have HA. if your create your own cache (and don't use the 'default' one) you can create it with Secondarys = 1 (HA on)
now let’s say you have 4 cache hosts; when you define a region , it will have both primary and secondary hosts. so each action on the region will result it being applied in both.
Shany
Named caches distribute across participating nodes. Named regions live on a single node. Regions can be HA, but they cannot take full advantage of distributed cache scaling, as their object load does not distribute across participating nodes in the cluster. Also, using named caches with HA requires three nodes minimum, rather than two nodes if you used the "default" cache only.
Related
So, I'm designing a distributed system with multiple redis instances to break up a large amount of streaming writes, but finding it difficult to get a clear picture of how things work.
From what I've read, it seems that a properly configured cluster will automatically shard and redirect requests made on the 'wrong instance' ( say key 'A' maps to instance 1 but is set on instance 2, it will be redirected to instance 1 ) Am I correct in assuming this?
If so, what advantages does an extra proxy and/or library cluster support give me over simply just connecting to one redis instance and letting it do all the work of figuring out where the SETS and GETS should be done?
Cluster support on the client means the client learns where the data is stored and remembers it, next time it tries to read or write a key it goes straight to the correct instance, which improves performance.
Its like calling directory enquires first every time you want to call a business vs just knowing the number of the business.
In case, I have my applications running in 2 different regions. If I use a traffic manager to route my packet to the nearest region, from a high availability perspective, how do I manage the manage data replication between these regions?
For example, if the application server in region A fails all my traffic will be routed to the application server in region B. Does this mean I will need bi-directional replication across both regions?
The short answer to your question is Yes.
for example we have three layers web,application and database.
We can create two web and two application servers with different location and can use SQL always on.
This way we can use traffic manager to point to two different regions.
If you need data parity between the two regions, you need to implement that since TM operates at the DNS layer and is agnostic to your data model. What it will do is direct your traffic to the best healthy endpoint in terms of latency (assuming you are using the performance routing mode).
Ok, so what I have are 2 web servers running inside of a Windows NLB clustered environment. The servers are identical in every respect, and as you'd expect in an NLB clustered environment, everybody is hitting the cluster name and not the individual members. We also have affinity turned off on the members in the cluster.
But, what I'm trying to do is to turn on some caching for a few large files (MP3s). It's easy enough to dial up a Redis node on one particular member and hit it, everything works like you'd expect. I can pull the data from the cache and serve it up as needed.
Now, let's add the overhead of the NLB. With an NLB in play, you may not be hitting the same web server each time. You might make your first hit to member 01, and the second hit to 02. So, I'd need a way to sync between the two servers. That way it doesn't matter which cluster member you hit, you are going to get the same data.
I don't need to worry about one cache being out of date, the only thing I'm storing in there is read only data from an internal web service.
I've only got 2 servers and it looks like redis clusters need 3. So I guess that's out.
Is this the best approach? Or perhaps there is something else better?
Reasons for redis: We only want the cache to use in-memory only. No writes to the database. Thought this would be a good fit, but need to make sure the data is available in both servers.
It's not possible to have redis multi master (writing on both). And I might say it's replication is blazing fast (check the slaveof command of Redis).
But why you need it in the same server? Access it as a service. So every node will access the actual data. If the main server goes down, the slave will promptly turn itself into a master.
One observation: you might notice that Redis makes use of disk in an async way. An append only file that it does checkpoint depending on the size from time to time so.
I'm a newbie to Redis and I was wondering if someone could help me to understand if it can be the right tool.
This is my scenario:
I have many different nodes, everyone behaving like a master and accepting clients connections to read and write a few geographical data data and the timestamp of the incoming record.
Each master node could be hosted onto a drone that only randomly get in touch and can comunicate with others, accordind to network conditions; when this happens they should synchronize their data according to their age (only the ones more recent than a specified time).
Is there any way to achieve this by Redis or do I have to implement this feature at application level?
I tried master/slaves configuration without success and I was wondering if Redis Cluster can somewhat meet my neeeds.
I googled around, but what I found had not an answer good for me
https://serverfault.com/questions/717406/redis-multi-master-replication
Using Redis Replication on different machines (multi master)
Teo, as a matter of fact, redis don't have a multi master replication.
And the cluster shard it's data through different instances. Say you have only two redis instances. Instance1 will accept store and retrieve instance1 and instance2 data. But he will ask for, and store in, instance2 every key that does not belong to his shard.
This is not, I think, really what you want. You could give a try to PostgreSQL+BDR as PostgreSQL supports nosql store and BDR provides a real master master replication (https://wiki.postgresql.org/wiki/BDR_Project) if that's really what you need.
I work with both today (and also MongoDB). Each one with a different goal. Redis would provide a smaller overhead and memory use, fast connection and fast replication. But it won't provide multi master (if you really need it).
I currently have some different project that works on different redis instance ( consider the sample where I've 3 different asp.net application that are on different server each one with its redis server).
We've been asked to virtualize and to remove useless instances so I was wondering what happens if I have only one redis server and all the 3 asp.net points to the same redis instance.
For the application key I think there's no problem, I can prefix my own key with the application name , for example "fi-agents", "ga-agents", and so on... but I was wondering for the auth session what happens?
as far as I've read the Prefix is used as internal and it can't be used by final user to separate... it's just enought to use different Db?
Thanks
Generally and unless there are truely compelling reasons, you don't want to mix different applications and their data in the same database. Yes, it does lower ops costs initially but it can quickly deteriorate to scaling and performance nightmare. This, I believe, is true for any database.
Specifically with Redis, technically yes - you could use a key prefix or the shared/numbered database approach. I'm not sure what you meant by "auth" sessions but you can probably apply the same approach to them. But you really shouldn't... since Redis is a single-threaded process you can end up where one of the apps is blocking the other two. Since Redis by itself is so lightweight, just spin up dedicated servers - one per app - even in the same VM if you must. You can read more background information on why you don't want to opt for the shared approach here: https://redislabs.com/blog/benchmark-shared-vs-dedicated-redis-instances