m******t
发帖数: 2416 | 1
Windows at least until vista has horrible memory
management.
On my day job, at any given time every developer needs
to run 3-4 Java VMs, each taking up 1-1.5GB memory,
plus a browser and other typical office software. Those
who do so on windows are constantly on disk swapping.
I'm not exaggerating - if you hear the their hard
drives spin, you'd think they are all hosting busy p2p
nodes. |
|
N********n
发帖数: 8363 | 2
Stack-based VMs all have JIT to handle optimization. Go's advantage over
there is not that big. The main performance gain is probably from that it
doesn't have as much exception handling as Java, which is always a pain
in the butt for any serious optimization attempt.
Go gives up a lot for performance gain I think. There's no exception, no
inheritance and no generics. You cannot write large-scale apps in it. |
|
w*s
发帖数: 7227 | 3 hi, i'm running 20 virtual machines (linux UML) in linux,
each of them can access a public folder in the host linux.
So i don't need to copy these apps/libs to each UML.
so i start all these UMLs, they will automatically run an app
in the public fold (/mnt/host).
But i run into
stack smashing detected error.
My questions is,
is these a bug in the application/lib itself,
or i'm running out of resources as i have 20 VMs ?
Thanks ! |
|
|
c******g
发帖数: 63 | 5 用的是VM player 5.02。一开始硬盘设置为20G, 现在想扩成30G。Power off后在
Utility菜单中已经将大小改为30G，按照网上一些资料的说明，剩余的步骤便是在
Ubuntu系统内部增加disk partition的大小。但似乎中间结果和网上写的有些不一样，
导致无法继续。具体过程如下：(assume所有命令都在sudo下）
$ fdisk /dev/sda 然后输p，显示：
Device Boot Start End Blocks Id System
/dev/sda1 * 2048 39845887 19921920 83 Linux
/dev/sda2 39847934 41940991 1046529 5 Extended
/dev/sda5 39847936 41940991 1046528 82 Linux swap / Solaris
输出n建一个新分区，显示：
Partition type:... 阅读全帖 |
|
k***r
发帖数: 4260 | 6 I had some problems with apps on a 64bit XP with 8G memory.
To be honest i find that 3G is pretty much all I need for
a desktop dev machine. So all my personal machines are
running 32 bit OS. I sometimes run VMs to take advantage
of my 8G memory at work but they aren't really that performant
even with a lot of RAM. I can't think of any other way to
use the RAM :) |
|
c*****t
发帖数: 1879 | 7 32-bit OS can work with 1 VM (at least in my old setup) at most
and 2 VM is just excruciatingly slow.
With work machine 12GB ram, quad CPU, I run 3 VMs pretty smoothly
(3 VM because I need to run hours long tests for one build,
2 other VM for current builds, one Linux and one Windows).
At the same time, I am still able to run office, VS, eclipse
(for writing small Java utilities, database client, and remote
GDB), browsers, and remote desktops. Currently total using
about 8GB, but it is nice not |
|
s*****g
发帖数: 5159 | 8 You will need the process run at different machines with IPs.
One (not simple) suggestion is to install two virtual machines on your
laptop, and manually assign IPs to the two VMs.
Afterward, starting on one vm, do
mpirun [your configuration file with two IPs specified] program
and see how it works.
I do feel debug on the cluster is a easy way for me - ssh in a terminal
will do most of the jobs. Do NOT submit jobs, just mpirun them from the
file / submission server and specify two nodes. Some cl |
|
j****i
发帖数: 305 | 9 Thanks for your reply. Next question is, how do I install two VMs, what
program do you recommend?
I don't want to debug on my cluster is because compilation is much slower
than on my laptop. To compile my code, it takes ~8 min on my laptop, but ~
20min on the cluster!
Thanks |
|
r*******n
发帖数: 3020 | 10 我做一个小的android app,在arm11的开发板上跑的挺流畅的.
我认为它跑的快主要是因为android的java程序跑在Dalvik VM上,
Unlike Java VMs, which are stack machines, the Dalvik VM is a register-based
architecture.(from wikipedia)
我觉得应该是上述原因,android app才运行的快 |
|
w*s
发帖数: 7227 | 11 not familiar with sockets, this is the info,
server talks with multiple clients, each client is running in an embedded
system.
when starts, server and clients can quickly make the connections.
now we use VMs to simulate the clients, all these "clients" run in 1 pc,
still talk with the remote server.
the thing is the connection time takes much longer.
why is this ？ if these clients use different ports, can the connection be
faster than all use the same network ports ?
thanks ! |
|
w*s
发帖数: 7227 | 12 was < 1 min for all ~100 client connections,
now > 10 mins,
actually we tried another way, 1 pc running 50 instances of client
simulation,
without VMs,
each client just makes 1 socket connection to the server.
still > 10 mins.
i don't know the internal of the client code.
thanks ! |
|
c****e
发帖数: 1453 | 13 Scaleout solution is not as easy as creating VMs in EC2. The whole pipeline
has to be designed to work in this way from scatch. It might be realtively
easy to add frontend instances, but the datastore is the hard part. Also,
there are many things consuming lots of effort to make it work in a public
cloud, such as security, throttling, virtual network, Geo distribution...
It might just be waste of resource to migrate to EC2. |
|
k******a
发帖数: 2436 | 14 2x Dell U2410. I want to upgrade to 3x U2711 this year, due to the
increasing number of VMs and remote desktops. If you have eye strain you
need to adjust ambient light and brightness of monitor till it is just right. |
|
c****e
发帖数: 1453 | 15 有MSDN的不花钱可以自己拿来玩玩side project.
http://weblogs.asp.net/scottgu/archive/2013/06/03/windows-azure
•No Charge for Stopped VMs
•Pay by the Minute Billing
•MSDN Use Rights now supported on Windows Azure
•Heavily Discounted MSDN Dev/Test Rates
•MSDN Monetary Credits
•Portal Support for Better Tracking MSDN Monetary Credit Usage
一个月有100-150 credit, 所有server一个价钱，才6c/h, 包括原来2块多钱的sql
enterprise. |
|
d*******r
发帖数: 3299 | 16 请问板上的 Python 大牛们 (比如公孙大神), 用 Python 做cloud上的VM管理 (比如
AWS 上大量 EC2 instance 的配置和monitoring), 主要需要用熟哪些工具?
我研究了下, 关于auto deploy/cfg 的工具, 选择有 SaltStack 和 Ansible.
据说是 Ansible 更简单, 跑在 ssh connection 上, 而且Vagrant 官方支持.
SaltStack 需要安装 client 到被管理机器上, 而且 SaltStack 反馈速度更快, 跑在
ZeroMQ connection 上.
不知道 Zeromq 这种东西靠谱不, SaltStack+Zeromq 到底坑多不多?
看 SaltStack 官网, 现在也有 Salt Air 了, 也可以模拟 Ansible 那种 ssh style,
被管理的机器只要能被 ssh 就行.
python 用来监控和管理 process 的最好工具是什么? supervisord?
还有, 可以用 python 的某些个 module/lib 来完全代替linux... 阅读全帖 |
|
n****1
发帖数: 1136 | 17 Js is the assembling language for the web. We need it, but it's painful to
write it with your barehand. However many language can be compiled into js,
including C++/java and even haskell.
I think it is technically infeasible to bring dartvm into chrome if google
still need v8 on chrome. It is nearly impossible to allow two VMs to
manipulate the DOM at the same time, because a) DOM is NOT thread-safe, b)
security will be a mess
Even PPAPI plugins need a JS bridge to communicate with DOM. |
|
c****e
发帖数: 1453 | 18 MongoDB is very flexible and has very little learning curve if the developer
knows RDBMS. Born with JSon + integrative query, and setup cost is almost
none but sharding costs your tons of VMs. Perf/scale wise, it can hold your
product/service very long without too much tuning, if you are in a startup ,
as least it will be OK after your A round fund raising, so you don't need
worry too much as at that time, you can hire more talented engineers to fix
the problem. |
|
p*****2
发帖数: 21240 | 19 http://queue.acm.org/detail.cfm?id=2567673
信息量太大了。
We did some prototyping with Ruby and Python using the evented frameworks
EventMachine and Twisted. The bottom line was that Node proved to be 2-5
times faster than both of those in terms of raw throughput. What was even
more exciting and really sold us on Node was that it took only two or three
hours to write the Node prototype while it took us more on the order of a
day or two to write the EventMachine and Twisted ones, just because we had
to ... 阅读全帖 |
|
d*******r
发帖数: 3299 | 20 我倒是觉得不要太依赖 AWS 那些独特的服务，不然就被lock住了。
我觉得只需要学会自动 create/terminate, start/stop VMs 就行，这些基本操作是每
个 cloud 的provider都支持的。保证你自己做东西的时候，不被任何provider lock住。 |
|
c****e
发帖数: 1453 | 21 When you think you are ok with "no join", it might NOT be true when product
evolves. Collection level lock is on the list for very long time, they
should have support that from day 1. MongoDB is very easy to hook up but it
doesn't scale well. I haven't seen any big footprint installment around.
Once you go with multiple sharding, it costs too many VMs.
dbs
table |
|
Y**G
发帖数: 1089 | 22 http://www.linuxjournal.com/content/containers%E2%80%94not-virt
Cloud infrastructure providers like Amazon Web Service sell virtual
machines. EC2 revenue is expected to surpass $1B in revenue this year. That'
s a lot of VMs.
It's not hard to see why there is such demand. You get the ability to scale
up or down, guaranteed computational resources, security isolation and API
access for provisioning it all, without any of the overhead of managing
physical servers.
But, you are also paying for lot o... 阅读全帖 |
|
L***s
发帖数: 1148 | 23 node做web server是合适的，因为不同用户的请求之间基本相互独立，
单线程异步的模型非常合适。
比如贱厂的web，一个页面大概包含3-6个micro services，每个service在
10-30个VMs上跑，每个VM大概起60-120个node进程，每个进程跑同一个app。
单个进程崩溃、重启、又崩溃、又重启，是家常便饭，不影响整体availability。
跟DB/storage打交道的逻辑，一般不在web server上做，是后台独立的，
跟web server之间用RESTful API隔离。
actually |
|
h**********c
发帖数: 4120 | 24 suggest you read Java Performance, written by Oracle engineers.
My thinking is GC should be handled out of the vm box. Single VM should be
stress-tested and the program/task should be able to throttle itself for
memory utilization and persist over usage to database. Use cluster/load
balancer to comprehensively conduct the VMs' symphony by major process
tables' reading or VM status readings -- process threshhold trigger balancer
audit and administration.
Unfortunately, in real life, we rather hav... 阅读全帖 |
|
g*********e
发帖数: 14401 | 25 Docker
The Docker Engine container comprises just the application and its
dependencies. It runs as an isolated process in userspace on the host
operating system, sharing the kernel with other containers. Thus, it enjoys
the resource isolation and allocation benefits of VMs but is much more
portable and efficient.
似乎他需要做每个container里面所有bin/lib 和 local OS的衔接，这个不好做吧？比
如linux 里面用fork() exec(), windows里是createProcess(), docker是不是都得安
排好这些？ |
|
c****e
发帖数: 46 | 26 说到akka这种mailbox方式异步通讯的模式，其实和几十年前OpenVMS（以前不叫这个名
字）的mailbox通讯方式很像，可惜自从DEC挂掉（挂掉的原因就是鄙视大众化的pc，而
且engineer culture导致上层还没法做出正确决断--这个和一些比较“潮”的程序员鄙
视java是不是很像？)，VMS都没人提了。 |
|
c********1
发帖数: 5269 | 27 I am running unbuntu 14.04, centos6.6, centos7.1 vms at the same time. My
computer have 32g memory. They are all ok.
I posted on unbuntu 14.04.
I allocated 6.4g mmmory for unbuntu 14.04. |
|
A***e
发帖数: 130 | 28 还是不太行。现在换了一台好点的机器：mbpr mid 2015, 16GB.
不过现在仍然是同一台physical host (client & server on different VMs). 需要两
台物理机器么？
见附图。 |
|
S****n
发帖数: 46 | 29 VIM is an improved version of the editor "vi", one of the standard text
editors on UNIX systems.
VIM adds many of the features that you would expect in an editor:
Unlimited undo, syntax coloring, split windows, visual selection, gui
support, and much much more (please read the text on reasons to use a vi
clone).
VIM runs on many operating systems:
AmigaOS, AtariMiNT, BeOS, DOS, MacOS,
OS/2, RiscOS, VMS, and Windows (CE/95/98/NT4/NT5)
and, of course, on UNIX in a lot of fl |
|
l*l
发帖数: 225 | 30 Normally, in the OS such as VMS or Multics ( first UNIX in 1970's ),
you can use detaching and attaching to get tge process which lost from
/dev/tty.
But POSIX 1.doesn't support connect with a lost term( /dev/tty ). But
several software can do this job.
you can try " screen " or "pty" to get your contral of the jobs. |
|
r*****r
发帖数: 2 | 31 【 以下文字转载自 Software 讨论区 】
【 原文由 rocktor 所发表 】
有个巨简单的问题：SET SERVICE怎么用？ |
|
