[THIS THREAD WAS MOVED AND IS A RESPONSE TO: http://www.satsupreme.com/showthread...371#post996371]
@Skyfly:
- Loopback... - Does not exist in cache-ex. You HAVE to use ONE group. Adding multiple groups.. will ruin oscam's de-duplication process, and multiplies load per attached group.
- cccmaxhops = -1 is not for any hidden sharing.. its to prevent.. sharing.. So having cccmaxhops = -1 is a good thing, not a bad thing..
Keep this in mind..
Thx!
hello m8,
I am very interested in a few things you are saying...
a very simple test proves you're wrong about loopback, let's set an example like this:
lets say you and i exchange cache, and for an example i have a large cache size, and you have only few peers and low usage, so rather low cache size;
when we exchange cache, you're cache size raises on a certain level and i see rather nice cache flow from you...but that's before i start testing...
then i do a simple trick and DISABLE YOUR READER, and check your cache size then...so the answer comes itself...if your cache decreases significantly, the logical conclusion would be i am receiving my own cache, that is my cache + my other peer's cache...(and i saw that very very very often...)
i had a lot of stability issues before i saw that, and removed a lot of peers that had that kind of config, and my oscam server is rock stable for as long as i am using this kind of configuration...
aside from that, what's the use of receiving my own cache, i already have it...
and about multiplying load by added group, i agree that is what happens, but please, i currently have 34 cache peers, that's 33 groups and my load never reaches 10% of cpu usage and 100MB of RAM used...(i am talking about stable VPS hosting and XEN virtualisation type)...
i agree that it can cause very heavy load, but it can happen on a crappy hosting and KVM virtualisation type (run away from KVM)...and of course on a PC with similar configuration...but i can't use it for an example...
and finally i completely agree with you about cccmaxhops = -1 being a good thing, but i find it "ugly" to see all readers "connected", and only a few with "needinit" status; i especially stated in my first post that i will not abuse anyone's reader, and expect the same from my peers, any of my peers can report any abuse from my side if being done ever
that would mean that i trust my peers by not "hiding" cards from them and expect the same from them, i don't expect that everyone would understand that, but don't see any normal reason why not to...in the end, again, i want to share with serious peers that i can trust, and want to be the person that they can trust to...
One simple test that proves - cache loopback still exist.
Receiver connected to server with reader and user account in same group. No other lines in receiver.
Server receive own cache from receiver.
Smiles... Thats a lot of text to answer to.. ;)
--
There are a few very long-lasting misunderstandings about cache-ex, and this indeed is a very important one.
Do a more simple test..
Connect your cache-ex oscam with another empty oscam (bi-directional), and on this empty-oscam put user + reader in same group 1.
=> As a result, the empty-oscam will receive your cache size obviously.
Now goto the CACHEEX page of your cache-ex server.. And check for incoming cache from that empty oscam..
=> There will be none.
What are the other effects of putting each peer in a unique group;
--
- The result will be tons of double cw's in your pool, which take more time to clean-up by the garbage process.
(=> while the garbage process runs, the cache is locked and does not flow).
- The flow-speed of oscam's cache-streams are highly affected by the total amount of cache (cachesize).
When putting each peer in a seperate group, no deduplication can take place, because this occurs.. per group.
- Your cachesize will be incorrect. Two peers with a size of 1000 each, will give 2000 total size in your server.
But in fact.. it isn't.. a lot will be duplicates.
- And mayb most important, you send this big pack full of duplicate cache to your peers..
Overloading their (hopefullycorrectly configured) de-duplication process.
Then about cccmaxhops...;
--
In my opinion its not a good idea to ask people to remove the -1.
CONNECTED may be the status for camd peers, but in CC-Cam it will reveal your card CAID's/shares and amount of peers.
Besides the fact that i not want to share that, its also technically possible to pull ecm's from cards without the -1.
You can trust your peers in full to not-do-that, but one mistake from a big server and your card/peers can and will be (ab)used.
--
Then.. Finally.. :)
About load. Yes.. putting all in one group takes CPU.., mostly because of the deduplication process/garbage collector with goal to create a 'pure' stream for your peers.
I have done extensive testing on this with oscam dev's aside (and also on all of the above).
For a mixed-CAID cache server with about 15000 (de-duplicated) cw size, you will almost fully occupy the cpu of a dedicated 6 core i7-3930k...
Yours.. is lower.. since it does not what it should do.. at the moment.
--
Try the test...
Just did your test here on my receiver.Originally Posted by nikko
Both cache-ex user and reader in group 50 (mode-3/cc-cam) and connected to my cache-ex..
Only the local decodes are sent back.. not the mass.
Exactly that I made for test empty-oscam in dm800 get cache from server and back same cache to server.CACHEEX page in server for client (dm800) must be none, but not.Server shows received cache even there hits from dm800!
Receiver all the time is on standby.
Reader dm800-cachetest 192.168.1.61 11318886237953363338X REVERSE CACHE PUSH 578172 0 0 0
Client dm800 192.168.1.61 11246828643915435402X REVERSE CACHE PUSH 0 52391 15 0 19
What protocol do did you use during the test ?
Booth in CCcam protocol.
Returned cache is less because of hops restriction.With larger hops returned more % of cache.
we are talking about two different things, but never mind, the fact is i had problems while my configuration was set according to what you say is correct, and now is ok, on my skyuk dedicated server i have 95% pure cache and forwarding it to my peers, so far no complaonts from any of them...
there are many peers from this forum, they can confirm what i am saying, as well as that we all had problems until we setup separate group on our readers...
so i will agree with you, there is no cache loopback on oscam, but there is something very similar to that, and we only have to come up with another name for this kind of behavior...
Well.. I would really like to know.. So many times asked.
Because in my test here nothing loops back.., while on Nikko's test it does...
And that's strange, isn't it..
Can somebody else do this simple test mentioned earlier ?
Thx
Posting Permissions
Reply With Quote