Just a little corner i asked for to write down weird and wonderful things
This is prob all boring to you, and well i wanted a place i could write stuff down from me so that you could have a small incite into my madness lmao, only kidding. So yes, this is prob all boring to you but i come accross stuff when trawling winmx bits and bobs and i really should write them down, but never do, so heres my blurb of crap, check back often to read blurb as an when you feel, its only for you admin folk anyhow, not meant to be a public page.
Cache design:
user sends request
|
cache recives 10 nodes
|
user attempts to join network using the 10 nodes.
How to optermise this process lies the question,
Current system as i know it sits connected to three secondrys, on request of user cache requests nodes from the primary relays it on to users,
a better way, to improve conenction times for all softwares, Robomx wont have updates, doesnt look like a new patch is on it way, new softwares im unsure if they will implement KMs mangle. this means that their will be a broad range of softwares about that will require either
A. KM's Mangle
B. Normal connection
Current cache softwares seem to return more lists of 10KM mangled patched nodes with on the odd ocascion you will get a unpatched node and robomx and hostfile users will connect, but times are lengthy, 30-40 secs at best. The new client really shouldnt use this mangle as we should come away from it, removing its crap from the network. But here in lies the problem are maintaining all softwares to keep folks intrested.
Suggestion,
Cache allows use from all softwares, how to do this, in the way the software index's.
Cache connects 3 secondrys.
|
Cache pulls nodes from these secondrys and stores them
|
Cache as it stores them sends basic send/recive to know its live or removes ip if fails
|
if its live cache stores ip using these rules
|
1. if 0 primary slots drop, not connected correctly to network
|
2. 1 connected primary is top of list for primary connections this is
|
so you get a good rotation of users round the network
|
3. 0 secondry and 4 primarys are added to top of list for secondry requests
|
cache stores 100 ips to create its own nodes list it sorts threw the list (index) to do two things, check that all 100 ips are still active, anymore is over kill and a waste of BW when indexing. organise the ips into patched and unpatched ips. splits them say into 50/50, if one side lowers on numbers during index the cache keeps making requests to its connected primarys(the 3 above) till it fills both patched/unpatched chambers so to speak
|
1.user makes request for primary conenction, cache returns 10 ips, 6 from patched, 4 from unpatched. (this can be changed during compile or config as the network maybe comes completely off the patch)
This should allow the fastest possible connection times for patched and unpatched softwares
2. user makes request for secondry connection, cache again returns 10 ips. and again 6 patched, 4 unpatched
Using the above example should maximise connection times for all softwares. It would also allow hostfile users to once again use our caches and old chat softwares/robomx etc to connect at lightening speed. No need for separate caches etc running diffrent softwares.
_________________________________________________________________________________________________
the next spiel lol
©2005-2024 WinMXWorld.com. All rights reserved. Page last updated Fri Jun 06 2008