s*******s
发帖数: 1250 | 1 windows 2012 r2 datacenter with many VMs |
|
a********c
发帖数: 3657 | 2 海门企业级。。。用的很少。。。datacenter基本是wd se的天下 |
|
z*********e
发帖数: 10149 | 3 有可能，也可能是产能上来了
今天早上手机新闻看到标题说他家datacenter产品的销量猛增 |
|
u**********e
发帖数: 282 | 4 It's not truth. Some users (gov., military, finance,etc) have certain
internal network for specific purpose (Storage, datacenter, etc). They won't
allow even port 80 opened. (One of our customer happen to be in this case)
I already found the solution from web. You can see more explanation here:
http://www.javacoding.net/articles/technical/rmi-firewall.html
from
port)
server |
|
|
m******t
发帖数: 2416 | 6
megarepository/
The article has been updated:
"... yottabytes are only one possible estimate for total storage
requirements. The more realistic estimates are
in the hundreds of petabytes, which is much easier for a datacenter to
accommodate. "
Still pretty impressive though. And I like 'yottabyte". It somehow makes
me think of "a lot of yogurt". |
|
v*****r
发帖数: 1119 | 7 Centos 的企业用户base其实蛮大的，举个例子，工作过的一家公司在 Datacenter 有
一百多台 dell blazer, 开始全 run 的 CentOS, 一个 SA 管，后来这个SA去了Google
，接管的两个SA 水平差点，对自己不够自信，把百多台 CentOS servers 都 migrate
了 RHEL。去 google 的那位 SA, 做 project 的时候首选还是 CentOS.
只要 RHEL 还在企业用户中 popular， CentOS 就不会 gone, 这是需求决定的 |
|
mw
发帖数: 525 | 8 大部分重要的服务器都是在远方的datacenter里面，你不用terminal用啥呢？ |
|
a9
发帖数: 21638 | 9 datacenter里上光纤的多还是用网线的多啊？
你可 |
|
c****e
发帖数: 1453 | 10 Windows is definitely catching up in this area. We have datacenter running
over 10K machines on windows.With windows server 2008 R2 and hyper-V, the
story could change in a few years. |
|
t****t
发帖数: 6806 | 11 昨天看maximum pc上介绍的facebook的datacenter, 说每天有四卡车的机器运来, 每卡
车13个rack, 每个rack 90个机器, 每个机器12核, 好多机器啊. |
|
c****e
发帖数: 1453 | 12 Faceboo很可能放大炮，buildout一般来说不会每天。Vendor到Datacenter调机器网络
都是有安排。而且这个每个rack才90个机器也挺少的。Rack里面的机器通信带宽大，太
小了不合适。不过FB可能也没有太多的computing需求，都买来做storage放照片了。 |
|
g*****g
发帖数: 34805 | 13 来自主题: Programming版 - 好热闹啊 For most companies, running their systems in cloud is probably more secure
than using their own datacenter. Think of it this way, how many companies
have highly qualified security experts? If not, it's more secure to use
those public clouds that have been designed by security experts and tested
by various companies. It's like keeping your stuff in the bank with others'
stuff next to yours, vs. keeping your stuff at home, hide it the best way
you can. |
|
c****e
发帖数: 1453 | 14 I am not sure about the main business of your company. I think Netflix
consumes huge amount of bandwidth as its main business is streaming. This
posts more requirements on the construction of datacenter, location(related
to US network topology) and CDN set up. I bet they did some math before
making the bet on AWS.
Other companies, let's say even twitter, might be more flexible on this.
of |
|
c****e
发帖数: 1453 | 15 Of course dev can be good admin as well. DevOps setup can even force them to
write better automation tool since they will feel the pain at first hand.
But when your org grow at certain scale, it becomes a big overhead to run
those repetitive tasks by dev team. No matter how good you claim your system
is, it always a dirty job to run infrastructure. Some one might feel good
to be a Decathlon winner if he is running a small hosted service. How about
60 clusters in 6 datacenters across the continen... 阅读全帖 |
|
g*****g
发帖数: 34805 | 16 Our team have 8 developers maintaining 7 clusters and counting in 6
datacenters. If test clusters are included, it's probably about 20 total.
It's one click on Jenkins to do building, baking application to AWS AMI,
deployment to a new cluster and bring down the old cluster. Alerts are built
into infrastructure. We can have some admins but their jobs would be only
passing on the alerts. I think the cloud computing has reduced the number of
admins needed. Instead of installing a system, deploying... 阅读全帖 |
|
g****t
发帖数: 31659 | 17 靠,您老泄露太多公司机密了吧.
Our team have 8 developers maintaining 7 clusters and counting in 6
datacenters. If test clusters are included, it's probably about 20 total.
It's one click on Jenkins to do building, baking application to AWS AMI,
deployment to a new cluster and bring down the old cluster. Alerts are built
into infrastructure. We can have some admins but their jobs would be only
passing on the alerts. I think the cloud computing has reduced the number of
admins needed. Instead of installing a ... 阅读全帖 |
|
d*******r
发帖数: 3299 | 18 满世界的新闻都是OpenStack, 二爷说说这东西到底牛在哪里，专门用来建datacenter
的吗 |
|
p*****2
发帖数: 21240 | 19
datacenter
牛在可以管理各种hypervisor吧？用户们一般不喜欢只用一家的hypervisor。 |
|
g*****g
发帖数: 34805 | 20 If you are rich, you can hire the best chef in the world, and they cook
better than most restaurants. On the other hand, 99% of families don't cook
as well as a chef in an average restaurant. That's why they still go out
dinning.
Security is like cooking. Only selected few companies can hire best security
experts to do in-house solution. All the others are better served in cloud,
and the cloud company can provide the security solution for you.
All accesses are remote, I don't see the difference ... 阅读全帖 |
|
c****e
发帖数: 1453 | 21 瞎吵吵之前，先想清楚定义。现在的背景下，底层的涵盖也在扩大。OS算，driver算不
算？做file system replication算不算？ cloud storage算不算， 做virtulization
的算不算？
这些的package非常高，也到处缺人。至于机会，系统的工作永远比应用的少。同时找
人的话对于经验的需求也高一些。说几个底层的例子：
（1） 给你1000台机器，怎么弄个dynamoDB出来？ 就算是在mysql上partition攒出来,
没个几年经验入不了门。
(2) 网络速度太慢了，rack机器可以用fiber直接connect,怎么自己优化protocol。
(3) 运算太慢了，要用专门的芯片协处理。为了在线更新方便，用FPGA。一个
datacenter里面的机器上这个，比run hadoop job有意思吧。
做什么都行，做自己擅长的事情。如果水平一般随大流的，奔着热门的找没错，
openning至少多很多。 |
|
p*****w
发帖数: 429 | 22 我支持你的观点，提个醒这些更偏infra不是底层。
瞎吵吵之前，先想清楚定义。现在的背景下，底层的涵盖也在扩大。OS算，driver算不
算？做file system replication算不算？ cloud storage算不算， 做virtulization
的算不算？
这些的package非常高，也到处缺人。至于机会，系统的工作永远比应用的少。同时找
人的话对于经验的需求也高一些。说几个底层的例子：
（1） 给你1000台机器，怎么弄个dynamoDB出来？ 就算是在mysql上partition攒出来,
没个几年经验入不了门。
(2) 网络速度太慢了，rack机器可以用fiber直接connect,怎么自己优化protocol。
(3) 运算太慢了，要用专门的芯片协处理。为了在线更新方便，用FPGA。一个
datacenter里面的机器上这个，比run hadoop job有意思吧。
做什么都行，做自己擅长的事情。如果水平一般随大流的，奔着热门的找没错，
openning至少多很多。 |
|
z*******3
发帖数: 13709 | 23 今天看了rod johnson的演讲，关于scala的前途的问题
http://parleys.com/play/51c1ffe7e4b0d38b54f46231/chapter16/abou
其实也没啥东西，它说的东西都在预料之中
但是不妨拿出来讨论一下，至少我觉得rod johnson这次做的这个演讲
把这个版面历次战争做了一个很好的总结吧
先从rod johnson这个人说起
你要是今天真的是写过java，靠java混过那么几个月以上的工资
你要是不知道rod johnson，那你简直就是活在中世纪以前的日本
基本上属于一个对挨踢世界无知的这么一个人
这个人，是澳大利亚人，悉尼大学的音乐phd，本科是cs的
它做出了一个spring框架，硬生生滴把当时ibm, bea, sun这些一干公司
给干翻了，当时sun的人出去给人做项目，他们自己都不用ejb
都在用spring，可想而知，这个人做出了一个多牛的东西
而实际上，用spring是我见过，可能从我接触计算机以来
最简单的一个东西了，我当时就用了五分钟，就把spring给用起来了
而且从那以后，从来没有说不懂spring过，太简... 阅读全帖 |
|
g*****g
发帖数: 34805 | 24 先说最重点的，availability你不能做。因为你是单机系统，负荷有上限，而我提出的
系统是没有上限的。另外hot standby，不能避免switch的时候交易丢失，你都在一个
机器上，数目很高，哪怕毫秒都要丢很多。再有，你这种单datacenter的做法，不能避免
网络出问题的availablity。LSE和nasdaq都出过类似的问题，导致系统关闭。而我提出
的系统，没有这个问题。
再说一下数据库划分依赖的问题，我做了最彻底的划分，系统有极大的余量，一天才几
万次写，
多100遍都不多。必要的话，当然可以把一些相关度高的线路放在一起。
最后再说一遍，你的一两台服务器，峰值要要负荷平时100倍以上的流量。而我的系统，
用的全是云上的VM，你买那两台机器的钱，就够我运营10年了。 |
|
g*****g
发帖数: 34805 | 25 Cassandra durability的东西是可调的。架构本身支持了写多Datacenter，丢数据的概
率是很小的。
普通数据库hot standby照样会丢数据，只不过丢得不多而已。我没有不允许你hot
standby。
我抓住不放的是内存数据库，当网络写流量大于硬盘流量的时候，你只能在内存buffer
，丢了就全丢了。内存是否够大也是问题。
同样的问题在MongoDB上大家都知道了。用得着你吹吗？ |
|
g*****g
发帖数: 34805 | 26 刚开始单机的，现在已经一串还复制到其他datacenter。我就请教一个问题，
假定一个DC着火了，另一个DC会take over，继续实时的订单处理对吧？
DC和DC之间想必是用ping来互相check了。我就想知道，两个DC之间网络断了一秒钟，
又好了，会怎么样？ |
|
v***e
发帖数: 2108 | 27 正派产品就一个，Couchbase Server，我们一直用
scalability比mongo好太多，还有mongo没有的
cross-datacenter replication（XDCR）和UI
get/set performance比cassandra强，因为是围绕
memcached 写的。
这些应该都是paid customer吧，不想花钱就用community
version好了，功能和enterprise version一样
当然，具体还要看你的需要 |
|
m********5
发帖数: 17667 | 28 我看到就想: what the crap!?
结果一看原始连接，发现果然是crap, 这个link是个过期的git tag，2012年的！
再看原作者当时怎么说：
http://www.reddit.com/r/programming/comments/v5ap8/web_server_b
I wrote the thing and I can tell you that yes, these results are pretty much
crap. I incorrectly used Folsom, the project I used to collect the stats.
Folsom is meant for realtime monitoring and not what I used it for. It only
keeps 60 second windows of the data it produces. That means all times prior
to the fine 60 second window was thrown out.
I twee... 阅读全帖 |
|
t**r
发帖数: 3428 | 29 如果是aware的话。
是说尽量把node跨rack还是尽量在一个rack? |
|
w**z
发帖数: 8232 | 30 当然不在一个Rack， 否则万一电源一断，或者switch 一出问题全完了。 |
|
