请选择 进入手机版 | 继续访问电脑版

Hi,Tokens

 找回密码
 立即注册
查看: 400|回复: 0

20180514井大和V神讨论分片等

[复制链接]

671

主题

1037

帖子

3788

积分

管理员

Rank: 9Rank: 9Rank: 9

积分
3788
发表于 2018-5-14 09:09:36 | 显示全部楼层 |阅读模式
文延:@天梓 @EOS原力 看了上面帖子说的EOS的所谓出块逻辑觉得很无助。这个不是什么提升呀
所谓的EOS超级节点纯粹是无稽之谈
传统关系型数据库做得都要比这个好,只有区区三十多个节点而已直接上mySQL/PostgreSQL搭一个上百节点的关系型数据库集群就可以达到百万TPS的性能,更不用说还可以有更高性能的DB2 Oracle Sybase

开心:@文延 不看好eos

文延:所以我不知道EOS想做什么,或者想给区块链社区带来什么,反正其实不需要有任何TPS高性能,主要有现在这么多人信,手工记账都可以大涨的,只要大家都信就可以,不知道为什么BM要给自己挖这么多坑。钱拿够其实就可以了。真实为大家担心到冒汗,现在的确是大投机,看谁运气好了。大家都在赌,看谁先崩盘

班长:EOS前四次提以太拉盘了,这第五次提以太没拉盘,很多人习惯吃进。结果150跌到90   不知道多少人站在山顶喝风,一下子超级节点奖励的30多个亿,韭菜们帮他掏出来了

文延:EOS在中国被极大吹捧估计是几乎没有中国人懂底层关系型数据库技术大数据技术或稍微懂一点的人没有入场,如果直接上国产的oceanbase, TiDB,HAWQ性能更是独步天下,我们看看EOS所依托的persistent layer其实就是chainbase,其实就是数据库技术中的玩具。大家玩得太神圣了,居然如此膜拜。炒作需要一点点常识为好。或者明明知道还炒也需要一点点良知和底线吧

文延:这点我想@Vitalik Buterin 最早是有先见之明的,所谓的DPOS就是一个非点对点的大型数据库集群而已,只有POW才是真正的点对点,这样的机制才是最安全的,也就是普通关系型数据库集群达不到的,理论上在我们做db2 v9.8的时候db2是可以设计成512个节点的,我自己亲自搭了128个物理机器的集群,都是可以支撑海量高并发transactions的,现在的所谓EOS即没有transaction management 的理念也没有任何data consistency所要求的数据一致性ACID的机制

文延:所以我的结论一直都是EOS现在其实是一场赌局,一定有人赚钱,有人赔钱,就看看谁是站着出来的那几个吧。有趣,我第一次买币ICO就是2017年参与了EOS走ICO流程,算是我也参赌了。拭目以待。看看这个剧本怎么写,大家又是怎么演

井底望天:@文延 eos运营模式有问题

文延:我还没有仔细研究EOS的运营模式,有什么问题?愿意倾听。Bitshare其实应该是EOS的纯净版,虽然也是DPOS但是无论如何还是有相对合理之处,现在EOS牛皮吹那么大,还不合乎逻辑。有理的牛皮吹大点实现不了大家还觉得水平不够,完全不合理的怎么去实现和圆谎

井底望天:@文延 你看看中本聪,在比特币系统起来的时候,他是假定系统面临最恶劣的环境
所以责任上是分布到每个小节点,而且留了60分钟给系统重新整合共识
虽然后来cpu,gpu,fpga,asic一直到大矿池
但是这是一个市场上不断试错,通过各种擂台打出来的结果,系统变得很强劲和稳定
eos的一上来就安排21个超级节点,已经把系统责任太多的集中在少数人身上
而这些靠钱来决定的少数人,有无技术保证,更为信念加持,大多为唯利是图之徒
再加上各种技术设计,都没有考虑环境的恶劣影响
这个系统能否跑起来,或者能否在黑客手上逃生,都是大问号

天梓 @EOS原力:EOS本来就没有颠覆什么,只是改进原有计算平台中心化严重的问题

天梓 @EOS原力:这个问号就交给实际运作来解答,任何区块链项目本身就是一场社会实验,成功失败都正常,但在问号没有变成句号之前,进行一番探索依然是必要的

天梓 @EOS原力:EOS只是给了区块链世界里对高TPS需求的一种解决方案,如果最后发现这个方案不可行,或者压根大家就不需要高TPS,那至少这个结论是由EOS这个项目得出的,以后的项目可以引以为参考

井底望天:@天梓 @EOS原力 非常欣赏你的社区
其实按照eos的白皮书,做出来也是可以的,不过大的技术方向,可能有问题
我对三个不可能,提出的方案,大家看是否有些意思


天梓 @EOS原力:@井底望天 谢谢

井底望天:@天梓 @EOS原力 也欢迎你们过来一起玩

井底望天:其实大家应该好好看看BM和Vitalik的技术争论,很有意思的
BM批评Vitalik,认为 1 简单分片,并不能解决不同的traffic之间的特异性 2 状态分片和跨片通信太多overhead
这些批评,是很中肯的
所以我自己就是先分层,把traffic分为底层的transactions和上层的smart contract
这样是上层分片,底层不分,避免在tranzactions这块,做避免双花这些麻烦问题

井底望天:但是Vitalik批评BM,认为不应该牺牲去中心化,也是对的
所以我是底层保持POW和去中心化
上层进行分片,采取POS,BFT,和各种用户自定义可插拔共识
算是目前比较好的一个解决方案
而且BM的可扩展性靠在超级节点加并行链,是采取了门槛不断提高的路线
我是靠增加小流量盒子,增加片的方法,是降低参与门槛,实践上应该更可行些

文延:@天梓 @EOS原力 “EOS本来就没有颠覆什么,只是改进原有计算平台中心化严重的问题” 刚好相反,EOS是通过增加中心化牺牲安全性来获得部分不牢靠的性能

gvn:The problems with multi-layered chain or sharding is that different chains/layers/shards have different security models with different strength/weakness, making bounded security assessment hard or close to impossible - and thus the system as a whole is most likely very brittle - there could be too many unpredictable attack vectors

Vitalik Buterin:that's not even how ethereum sharding worka
every shard has the same securitt
security

Vitalik Buterin:我们研究了分开交易和计算的概念很多
问题在于,正常交易的计算需求不那么复杂,只需要验证数字签名而已
如果只分片计算,不分片数据,处理很多交易还需要很高的performance requirements
比如,10万交易每秒:一个交易是250字节,所以一共需要下载25兆每秒
乘以p2p网络的overhead,可能需要1 Gbps的互联网联通
所以要选择:分片,还是中心化

gvn:@Vitalik Buterin So a natural bound of single blockchain throughput is a single node's network capacity (bandwidth)?

Vitalik Buterin:yes, if the blockchain is not sharded


艾迪:这是本人吗 还是代管账号


Vitalik Buterin:我是真正的vitalik,但是如果我是骗子,我也会这么说

gvn:From the quality of the discussion, it should be the real @Vitalik Buterin - a kind of proof of (mental) work
It's good enough for the sake of the discussion

井底望天:@Vitalik Buterin of course, sharding based on decentralived structure is preferred.
But  I think that multilayered approach is a good way to resolve the trilima.
@gvn our model,the microchian (shard) has same security level as base (mother) chain

gvn:But how do yo you mitigate the potential safty degradation in a heterogeneous multi-chain solution? @Vitalik Buterin just pointed out above that in his ethereum sharding solution "every shard has the same security", so is Bitcoin's sidechain - which tries to solve this problem by "shared mining"

gvn:of course Bitcoin's sidechain is not a sharding solution - but in general multi-chain, as well as sharding, are all plagued by this safty degradation problem

井底望天:@gvn because our micro chain (sub chain, child chain) is not side chain. The micro chain nodes communicate each through mother chain. So you cannot direct attack the micro chain. You have to attack the mother chain which is POW with over 10,000 mining nodes running.

gvn:@井底望天 yes this is shared mining - but sharding means that each shard does not have the full chain to do full validation so it relies heavily on the main chain - and it is not a problem in a heavily fortified datacenter but could be vulnerable in the wild open internet for a public blockchain environment

井底望天:@gvn not really true, we separated the smart contract traffic in the shards, an every
And every shards kept the only consensus and flush the states to the main chain periodically.
Every shard behaves like a logical full chain.

Vitalik Buterin:sidechains don't solve the problem
either all miners verify all chains, or each chain has only a few miners verifying it
ethereum sharding solves this with random sampling, fraud / data availability proofs and tight coupling

gvn:@井底望天 I understand - it is not about right or wrong - in distributed systems it's all about your presumptions about bounds - your snapshot   flushed to the main chain is not fully validated

井底望天:@gvn we didn’t use sharding as a pure technical solution but making it bounds the smart contract as a logical sub chain. It more like creates a new model for DAPP.

gvn:@Vitalik Buterin Not directly related, but something like VRF is needed to solve this problem

井底望天:@Vitalik Buterin how reliable with random sampling?

gvn:by VRF I mean Verifiable Random Functions -

兰杰:random sampling is a very solid way to solve a lot reliability issue of fault tolerance. The only drawback is random nowadays is not true random.

gvn:to achieve a high degree of true "randomness" some external entropy is needed to be injected into the underlying system "sampling" that random source - which fortunately is not a problem for either PoW or PoS systems, they both have external entropy directly parametered inside their protocols



本帖子中包含更多资源

您需要 登录 才可以下载或查看,没有帐号?立即注册

x
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

手机版|小黑屋|Hi,Tokens  |网站地图

GMT+8, 2019-10-16 00:54 , Processed in 0.053990 second(s), 4 queries , File On.

Powered by Discuz! X3.3

© 2001-2017 Comsenz Inc.

快速回复 返回顶部 返回列表