j*a
发帖数: 14423 | 1 root@:~# ethtool eth0
Settings for eth0:
Supported ports: [ ]
Supported link modes:
Supports auto-negotiation: No
Advertised link modes: Not reported
Advertised pause frame use: No
Advertised auto-negotiation: No
Speed: 10000Mb/s
Duplex: Full
Port: Twisted Pair
PHYAD: 0
Transceiver: internal
Auto-negotiation: off
MDI-X: Unknown
Link detected: no
root@:~# ethtool eth1
Settings for eth1... 阅读全帖 |
|
w**z
发帖数: 8232 | 2 CMS 也是有stop the world pause， 只不过它很短， 发生在initial mark 和remark
phase。 JVM 的一个误区，是觉得内存越大越好，其实内存越大，GC pause 越长。
Cassandra 就不建议每个JVM heap 超过8G.CMS 在heap 超过8G，会有问题。C＊尽量使
用off heap memory， OS cache etc. Kafka 也是相同的思路，尽量利用现成的 linux
pagecache.
其实对distributed system 来说，单机性能，并不重要，主要是做到horizontal
scalable.单机都是commodity hardware， 不行，就加机器。在这上，Netflix 做的就
非常好，那不是那么简单的。数以百记的service， calling each other， 数以万记
的machine， 不停的deployment， machine up and down, joining and leaving
pool.
你所说的 “ 但当面对12306的用户群时，巨大的访问量会给... 阅读全帖 |
|
b***e
发帖数: 1419 | 3 /**
To use promise is unnecessary. More succinct syntax can be achieve with
other libs, such as "node-sync". This package also captures an emulation of
call stack when there are exceptions.
**/
/* use "node-sync" package */
var proc = require('node-sync');
var cb = function(err, res) {
if (err) {
console.log('cb ERROR', err.stack);
console.log(err.__generatorStack__);
} else {
console.log('OK', res);
}
};
var wait = function(n) {
return function(cb) {
... 阅读全帖 |
|
|
