Jump to content

Recommended Posts

  • 3 weeks later...
Posted

New feature

 

Package installations:

  • Detects when its faster to use a specific package
  • Download many files in a custom tarball compressed with gzip
  • extracts only the needed data
  • detects first time installation to download faster a single package
  • 5 months later...
Posted (edited)

Working on this again, also decided to share a beta version so you can see how insanely fast this thing is.

 

Test Server

 

 

CPU: VPS 2vCore

SSD

Bandwidth: 100 mbps

HTTP daemon: Lighttpd

OS: Archlinux

 

 

 

Test Client

 

 

CPU: i7 6950X

SSD 850 Pro

Bandwidth 50mbps

OS: Win10 x64

 

File Scan: < 20ms

Full update: < 25 seconds

 

 

 

Beta client can be downloaded bellow, it will try to download a file patch of size ~= 120mb.

The executable is not signed and windows or chrome may warn you not to download it, but there's nothing to worry about Virus Total

 

The bellow client may go into non-working state any time because the instructions offsets may get updated

 

Download Link

 

The message Initializing...l means that the updater done it's work and waits for L2 client to launch but this feature is not yet implemented, just hit the Cancel button

Edited by xxdem
Posted (edited)

PS: ALL the files on this updater are automatically considered as "Critical updates" this updater in other words will force "Full Check" everytime its run. A "Full Check" scenario on an old-obsolete updater is about 5-8 minutes on an average patch size, this updater is able to use 100% of the available hardware resources and give's the ability to read/scan the same patch within 50-100 MS

Edited by xxdem
Posted

Update

 

After lots of work, succeeded onto making stream decompression n-memory while downloading. This feature is totally unique among all updaters and I will explain why.

 

Technical explanation of how it works

 

Old mainstream method

 

 

Download all bytes and append/write intto xxx.zip on the hard disk

Buffered reading of xxx.zip and decompressing in memory again

Stream de-compression of in-memory bytes back to the harddisk and appending into un-compressed file xxx.u

 

A total of 3 steps involving heavy IO operations

 

 

 

In memory (new method)

 

 

Download next available data chunk -> decompress in-memory -> append to xxx.u

 

A total of 1 step reducing IO about 66% in comparison with the old method. It won't bottleneck your download speed since it operates in different worker thread.

 

 

  • 2 weeks later...
Posted

Threading has nothing to do with bottlenecking the connection.

This connection bottleneck is caused because the download function is trying to use as much bandwidth as possible so it can be downloaded as fast as possible.

 

Isn't that correct?

 

In theory, if u put additional workload on the very same thread that downloads through a socket, you add latency on the connection eventually decreasing network efficiency

Posted (edited)

Isn't there a better way to throttle the bandwidth of the download tho?

 

There is no throttle involved during a normal download, the old version or even the standard updater we are all used to will download the file with the highest efficiency possible. But will then require to decompress the compressed file into a new file on the disk with it's own new size, this means delete the downloaded compressed file and decompress it into a new file, which I believe adds unnecessary IO.

My method is more simple, it will decompress the bytes as soon as they are downloaded from the socket while downloading, this method will introduce input lag on the download thread and will throttle the network down, so this was solved with a second worker thread that received job from the download thread, decompressing the bytes in a streamed writer who is also responsible of writing the decompressed bytes to the hardisk through another buffered writer

Edited by xxdem
Posted (edited)

your virustotal need update 

virustotal

 

yes its a virus, I want to scam your endless money

 

endgame antivirus

 

jvsIqgz.gif

Edited by xxdem
Posted (edited)

yes its a virus, I want to scam your endless money

 

endgame antivirus

 

really you are so idiot ?

i said only you scan need update nothing more ...

i'm sure you don't have the Knowledge to create anything like this ...

Edited by pirama

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now



  • Posts

    • First, don't really follow the "main voice", moreover if you consider it an hobby. Simply do what you want, you got only one life so use it as you want. If you make it an hobby, it's exactly like piano, or velo - only practice makes you better.   Secondly, how do you learn things ? It's actually a really important question, since some can simply be scholar, read books (theory) then practice ; and some simply can't read books. I'm the second type, I hated school, I find it boring - my knowledge in Java comes from try-and-fail. You improve your coding style every year or so, I can myself rewrite my own code (which I already considered top-notched) after a while. You always learn something new - even if Java barely evolves. L2J is a fun way to learn programming, it's a giant sandbox where you can edit anything, and I believe it should be taken as it.   My own way of learning was as follow : Add existing customs, no matter what they are : the point is to know main classes used by L2J / customs. L2J is barely Java knowledge ; the true knowledge is to know WHAT to search in WHICH location (what I call, organization). You have to understand than EVERYTHING you think already exists, in a form on another, in the source code. A custom is only the association of the different mechanisms you found "here and there", glued together in a proper goal. Once you know main classes to edit, and the customs you added are compiling fine, the main point is to know WHAT exactly you DID. Try to understand WHY and WHERE you actually copied the code. Third point would be to MANIPULATE the customs you added in order to fit your wish. First edit little values, then logic conditions ; eventually add a new Config, or a new functionality to the custom. Fourth point would be to begin to craft your own ideas. Once again, EVERYTHING already exists, in a form or another. You want a cycled event ? You got Seven Signs main task as exemple. Npc ? Search any type of Npc and figure out what it does. Fifth point would be to understand Java - mostly containers (WHAT and WHERE to use them), variables types and main Java mechanisms (inheritance, static modifier, etc). You should also begin to cut your code into maintainable classes or methods. Java can actually run without optimization, but bigger your ideas, more optimized and well-thought it should be. It's direct saved time in the future, and you would thank yourself doing so. Main tips : ALWAYS use any type of versioning system - GIT or SVN. It allows to save your work, step by step and eventually revert back anytime you want if you terribly messed up. L2J is 80% organization knowledge, and 20% Java knowledge. Basically, if you know WHAT and WHERE to search, if you aren't dumb, it's easy to replicate and re-use things. Cherry on top is to use a already good coded pack to avoid copy-paste crap and get bad habits. Avoid any type of russian or brazilian packs, for exemple - their best ability is to leak someone's else code. Obviously you need some default sense of logic, but Java and programming in general help you to improve it.   Finally, most of your questions could be solved joining related Discord (at least for aCis, I can't speak for others) - from the moment your question was correctly asked (and you seemed to search for the answer). My community (and myself) welcomes newbies, but got some issues with noobies.   The simpliest is to try, fail and repeat until you succeed - it sounds stupid, but that's basically how life works.   PS : about Java ressources, before ChatGPT, it was mostly about stackoverflow website, and site like Baeldung's one. With ChatGPT and alike, you generally double-cross AI output to avoid fucked up answers. Also, care about AI, they are often hallucinating really hard, even today. They can give you complete wrong answer, you tell them they are wrong, and they say "indeed, I suck, sorry - here's a new fucked up answer". You shouldn't 100% rely over AI answer, even if that can give sometimes legit answers, full code or just skeletons of ideas.   PPS : I don't think there are reliable ressources regarding L2J itself, also most of the proposed code decays pretty fast if the source code is actually maintained (at least for aCis). Still, old coded customs for old aCis sources are actually a good beginner challenge to apply on latest source.
    • WTS: - AQ - Baium - Zaken  - Frintezza - Vesper Fighter Focus Fire Element   pm for detalis
    • We have the best price! L2Reborn.org Signature x1 Franz NEW!! 1KK = $20 HURRY TO BUY AT THE TOP PRICE discord - adver745645
  • Topics

×
×
  • Create New...

AdBlock Extension Detected!

Our website is made possible by displaying online advertisements to our members.

Please disable AdBlock browser extension first, to be able to use our community.

I've Disabled AdBlock