m**k 发帖数: 290 | 1
hardware forwarding 没有问题。
hardware fragmentation 应该也没有问题吧,不是很确定。
hardware re-assembly 不需要router处理,所以router不需要在硬件里实现这个功能
。除非做 NATing 是需要的。NATing 和 router的硬件差别应该很大吧。
TCP支持 PMTU discovery. 当 TCP 发一个数据包比 pmtu 大时,这个数据包会被某个
router 分割,这个 router 会返回一个 ICMP unreach with reason "frag needed".
当TCP 收到这个 ICMP 时,会自动更新 PMTU.
IP 协议规定所有的 router 都必须支持 ICMP unreach. 所以即使 router 不进行分割
,或者不转发分割包,TCP 也可以正常工作。
UDP不支持 PMUT discovery. 常用的网络协议,比如 DNS, NTP 用的数据包都很小,不
需要分割。如果需要发送大 UDP 包,那就需要在应用层实现 PMTU discovery.
现在的 Int... 阅读全帖 |
|
m**k 发帖数: 290 | 2
hardware forwarding 没有问题。
hardware fragmentation 应该也没有问题吧,不是很确定。
hardware re-assembly 不需要router处理,所以router不需要在硬件里实现这个功能
。除非做 NATing 是需要的。NATing 和 router的硬件差别应该很大吧。
TCP支持 PMTU discovery. 当 TCP 发一个数据包比 pmtu 大时,这个数据包会被某个
router 分割,这个 router 会返回一个 ICMP unreach with reason "frag needed".
当TCP 收到这个 ICMP 时,会自动更新 PMTU.
IP 协议规定所有的 router 都必须支持 ICMP unreach. 所以即使 router 不进行分割
,或者不转发分割包,TCP 也可以正常工作。
UDP不支持 PMUT discovery. 常用的网络协议,比如 DNS, NTP 用的数据包都很小,不
需要分割。如果需要发送大 UDP 包,那就需要在应用层实现 PMTU discovery.
现在的 Int... 阅读全帖 |
|
L******t 发帖数: 1985 | 3 网管看到fragmented traffic会不会调整网络避免fragment?
什么use case下traffic fragmentation常见,或者说被接受? |
|
L******t 发帖数: 1985 | 4 网管看到fragmented traffic会不会调整网络避免fragment?
什么use case下traffic fragmentation常见,或者说被接受? |
|
g****t 发帖数: 31659 | 5 我说你们都没考过GRE么。
Fragmentation越多,貌似比单一系统工作position多。这有个隐含假设就是
fragmentation不会对Android的推广有影响。这些然是错的。
推广到极端,你弄十万个fragmentation.
系统自己就死了,而不是多造了十万倍的工作机会。
云计算几大平台完全不兼容,负面作用也是很明显的。
了, |
|
r******y 发帖数: 3838 | 6 One of the major talking points long used by Google in support of its
Android smartphone operating system
over iOS is its "open" nature that has allowed handset manufacturers and
others to tweak and customize the
software for their needs. Apple CEO Steve Jobs has argued that the "open"
nature would more accurately be
described as "fragmented" in justifying why he believes that Apple's "closed
" or "integrated" iOS is a better
platform for consumers.
Google executive Andy Rubin responded to Jobs'... 阅读全帖 |
|
m**k 发帖数: 290 | 7
ISP一般都允许 fragmented packets.
接入网可以丢掉这些包。因为edge router的firewall filter大都是stateless, 没有
办法对这些包分类过滤。丢掉这些包可以避免 flooding attack.
常用的网络协议都支持 pmtu discovery. 所以禁止fragmentation对应用层影响不大。 |
|
m**k 发帖数: 290 | 8
ISP一般都允许 fragmented packets.
接入网可以丢掉这些包。因为edge router的firewall filter大都是stateless, 没有
办法对这些包分类过滤。丢掉这些包可以避免 flooding attack.
常用的网络协议都支持 pmtu discovery. 所以禁止fragmentation对应用层影响不大。 |
|
l******u 发帖数: 1174 | 9 碰到一个问题,一个C++写的server,run了一天后,用top看占了1.5G,在它想再分配
一块500M的heap时,就out of memory了。
同样的程序,在另一个server上run了一天后,却只占300M,也没有out of memory的问
题。
本人怀疑是memory fragmentation。因为如果是memory leak的话,不管在哪个server
上都会有memory leak。可能两个server的kernel不一样(虽然都是RHEL3),导致只在一
个server有memory fragmentation?
哪位有这方面经验的能分享一下吗? |
|
w*********a 发帖数: 9279 | 10 【 以下文字转载自 CS 讨论区 】
发信人: wugongpanda (Sela'ma ashal'anore!), 信区: CS
标 题: 深受memory fragmentation毒害。少用长链表
发信站: BBS 未名空间站 (Mon Feb 25 11:46:31 2013, 美东)
c++分配内存的方法确实不好。分配内存效率低。
尤其是长链表,一旦free之后,fragmentation太多。严重影响效率。
而且关键是不cache friendly.
从这个意义上说,reference counting 也有这个不足。 garbage collection要好得多
。 |
|
m***o 发帖数: 1367 | 11 Does anyone know a good way to guess/tell/estimate heap fragmentation under
Solaris 2.x?
My app is slowly running out of memory, although the memory increasing rate
went down with time after several days. So I am not sure it is caused by some
small leak or heap fragmentation.
Thanks. |
|
x********u 发帖数: 430 | 12 Thanks. I will try that.
Some people suggested me to select some unique RE sites and clone part of
the fragment first, then comes to the other part of the fragment. I dno't
think this is a good choice. |
|
b******k 发帖数: 1874 | 13 so, if i am going to ligate two tandem short fragments (say, 50bp each):
synthesize 50nt oligo of Sense and antisense (with 5' phosphate group modifi
cation),
anneal them by boiling and cooling naturally,
then ligate the annealed fragments (ideally, with synthesizd 3' overhang fro
m first one, and 5' overhang from 2nd)by conventional T4 ligase system?
thanks. |
|
c*****e 发帖数: 619 | 14 I want to break oat genomic DNA into fragment of 10KB, what should be the
setting? So that I can get the fragment of the size I want?
Thank you! |
|
x********u 发帖数: 430 | 15 Is it possible that I insert a 6Kb fragment into a 8Kb vector? I have
repeated three times and have no luck geting any colonies. For construct
below 10Kb, I am very confident (>80%) and I can always get the right
transformants after screening. Is there any tricks for this big fragment
cloning? I know all my gene can be functionally expressed in E coli, so
there is no protein toxicity issue. |
|
a***y 发帖数: 19743 | 16 【 以下文字转载自 PDA 讨论区 】
发信人: aaaty (阿fan提), 信区: PDA
标 题: Re: Fragmentation of Android Phones
发信站: BBS 未名空间站 (Thu Oct 27 13:59:22 2011, 美东)
Besides lack of timely updates, let alone updates at all, customers are
generally not given any guarentee that their phone will be continuously
updated in a timely fashion. Customers are left to bet and guess. That will
definitely hurt Android platform retention rate.
Combined with most people do not pay for apps on Android, retention will be
even worse. |
|
w********r 发帖数: 14958 | 17 2周我苦思冥想阿。 debug的程序都快加遍了。
我这个mpi的程序要跑几天才会出错 segmentation fault。 而且每次还不一定什么时
候才出错呢。 今天终于发现是new 的太多了。 可是我感觉delete应该也很及时。
今天发现virtual mem已经用了都好几G了,正经ram根本用不了几兆。heap里面的
fragment太多了。 最后慢慢就挂了。 |
|
w*********a 发帖数: 9279 | 18 c++分配内存的方法确实不好。分配内存效率低。
尤其是长链表,一旦free之后,fragmentation太多。严重影响效率。
而且关键是不cache friendly.
从这个意义上说,reference counting 也有这个不足。 garbage collection要好得多
。 |
|
|
w****n 发帖数: 266 | 20 请问在考虑tuple-level security granularity的情况下,如何实现horizontal
fragmentation? |
|
L******t 发帖数: 1985 | 21 IPv4 routers虽然理论上都支持fragment oversized packets,但实际上hardware
forwarding做不了,software forwarding doable但肯定not desirable。所以我想知
道网管的common practice是什么?
关于常用的网络协议都支持pmtu discovery,能否举个例子?什么时候IPv4 Host在往
某个目的Host发包前会先做PMTUD? |
|
m**k 发帖数: 290 | 22 Back to your orignal question. Fragmentation is a sub-optimal case, and
should be avoided when possible.
For example, if you use tunneling or MPLS in your backbone, you should
increase your backbone MTU to accomodate for the tunneling headers. |
|
x*********n 发帖数: 28013 | 23 “常用的网络协议都支持pmtu discovery”这个不对。
fragmentation只在TCP讲,TCP上面有6个flag,其中一个bit DF就是给这个pmtu
discovery准备的。所以只要跑TCP,这个就会支持。
by default,一个包发过去,只看interface的容量,interface不到就会丢包
pmtu discovery junos以前版本都是disabled,后续版本才开始by default enabled,
很多别的厂家不用的。这个只有在别的端 device不确定的时候,才用,不然我改TCP
mss就好了。 |
|
L******t 发帖数: 1985 | 24 IPv4 routers虽然理论上都支持fragment oversized packets,但实际上hardware
forwarding做不了,software forwarding doable但肯定not desirable。所以我想知
道网管的common practice是什么?
关于常用的网络协议都支持pmtu discovery,能否举个例子?什么时候IPv4 Host在往
某个目的Host发包前会先做PMTUD? |
|
m**k 发帖数: 290 | 25 Back to your orignal question. Fragmentation is a sub-optimal case, and
should be avoided when possible.
For example, if you use tunneling or MPLS in your backbone, you should
increase your backbone MTU to accomodate for the tunneling headers. |
|
x*********n 发帖数: 28013 | 26 “常用的网络协议都支持pmtu discovery”这个不对。
fragmentation只在TCP讲,TCP上面有6个flag,其中一个bit DF就是给这个pmtu
discovery准备的。所以只要跑TCP,这个就会支持。
by default,一个包发过去,只看interface的容量,interface不到就会丢包
pmtu discovery junos以前版本都是disabled,后续版本才开始by default enabled,
很多别的厂家不用的。这个只有在别的端 device不确定的时候,才用,不然我改TCP
mss就好了。 |
|
m***i 发帖数: 7 | 27 DF is ip layer flag, so both UDP and TCP support ip fragmentation |
|
l******u 发帖数: 1174 | 28 1.5G is what the columns "SIZE" and "RSS" of the "top" command show.
I need to allocate buffers for incoming socket messages, some of them could
be huge.
The reason I suspect it is memory fragmentation is that the "out of memory"
error:
1) does not always happen on server A, only once every 5-6 days.
2) does not happen on server B at all, even though the cod base and input
are identical.
3) only happens for the largest 1-2 messages. Even after it happened,
small messages still work.
When |
|
h**********c 发帖数: 4120 | 29 tcp 怎么会有fragment呢?
有什么文献看吗?RFC?
有什么套路没有?
举个例子,还真碰到一次,
send 488 bytes
服务器受到484
再收到4
这是个cs问题,ce问题,还是ee 问题? |
|
h**********c 发帖数: 4120 | 30 用的openssl biosocket,参考上明确的说是tcp,
另外apache mina介绍codecfactory明确说过tcp会有fragments,但是讲得也不详细。
有什么参考介绍一下吧。 |
|
g****t 发帖数: 31659 | 31 android相关工作确实更多。这个没疑问。
但你还用得着琢磨job security?
另外虽然结论符合现实,你的推理明显有问题。fragmentation可能影响到android总总
用户下降。
吗? |
|
n****1 发帖数: 1136 | 32 以此类推, 云计算fragmentation也应该是好事. ec2/azure/gce/openstack做到完全不
兼容是第一步, 终极目标是ec2上的各种不同instance之间也不兼容.甚至是每一个
instance都不兼容.
如果维护本地机器和服务需要1个developer的话,ec2要3个。但ec2占有率都这么高了,
这个平台不能放弃,还有嫌坑多的吗?
如果oracle能彻底打破java一次编译到处运行的习惯, java程序员肯定开心极了吧. 反
正这世界是离不开java了 |
|
n****1 发帖数: 1136 | 33 我一向更喜欢android,只是我觉得缺陷就是缺陷,应该正视和解决, 而不是抱着轮子的
心理颠倒黑白.
你这番"fragmentation是优点"和果轮们的手随iphone成长是一个路子. |
|
g****t 发帖数: 31659 | 34 fragmentation少些的话,职位说不定更多。 |
|
n****1 发帖数: 1136 | 35 fragmentation直接结果是工作量变大, 但没有证据表明它导致android职位多,程序员
要加更多的班倒有可能.
软件这个东西本来最关键的就是控制复杂度, 同一个项目在保证按时完工的前提下,参
与的人数越少越容易开发维护, 反之则会因为风格/沟通/扯皮等因素恶性循环直到一发
不可收拾.
如果一个app同时推iOS和android版本, 大致的架构和逻辑是一样的, iOS版的一个人能
搞定, android版却要三个人写. 那三人都会是痛苦的, 也是可悲的.
如果我是老板, 我会选择炒掉两个,然后榨干第三个人的每一油水. 实在不行, 大不了
bug多点,用户体验差点,发布晚点嘛.绝大多数android app都这个德行,我为啥多花那个
冤枉钱. |
|
x********u 发帖数: 430 | 36 I am going to clone an operon with two genes, the total length is about 4500 bp. Is it possible that I clone this big fragment all at once and get the
right sequence? I am using the dual-promoter system and E.coli host strain. Many
thanks for your help. |
|
X***n 发帖数: 366 | 37 Try SLIC.
Nature Methods - 4, 251 - 256 (2007)
Published online: 11 February 2007; | doi:10.1038/nmeth1010
Harnessing homologous recombination in vitro to generate recombinant DNA via
SLIC
Mamie Z Li & Stephen J Elledge
4500 bp. Is it possible that I clone this big fragment all at once and get
the
. Many |
|
s********n 发帖数: 2939 | 38 设计oligo的时候可以把粘性末端设计好,链接前可以不用做酶切。
我试过的一个protocol,非常简单好用:
1. oligos用TE or H2O溶解,浓度100 uM;
2. 磷酸化准备如下,37C一个小时, 70C for 10 min.
15 ul H2O + 2 ul oligo (100 uM) + 2 ul T4 ligase buffer + 1 ul TNK
3. 退火反应如下,95C 5 min in PCR cycler,RT for 30 min.
18 ul H2O + 1 ul oligo F + 1 ul oligo R (final 0.5 uM fragment)
这个产物可以直接用于连接。 |
|
g*********5 发帖数: 2533 | 39 I think mutation is the easiest way to do. just with quikchange kit, about
200 bucks.
because your pcr fragment is smaller than 30 bp, so you could design primer
with this sequence and flanked with vector sequence, 15 bp each, then use
this 60 bp primer pair do insert mutation, just one PCR and transform.
good luck |
|
w********a 发帖数: 324 | 40 DNA fragment 合成,< 500bp;
除了IDT的gBLOCK 和Life tech的GeneArt® Strings™ 还有哪家提供这个
服务,而且价格比较便宜?
谢谢!! |
|
H****N 发帖数: 997 | 41 Anybody have experience of synthesizing large (~10 kb) DNA fragments? Which
company do you use? Thank you. |
|
H****N 发帖数: 997 | 42 Anybody have experience of synthesizing large (~10 kb) DNA fragments? Which
company do you use? Thank you. |
|
w***e 发帖数: 269 | 43 这里的意思就是anti-digitoxin antibody.至于Fab不fab是这样子的。antibody可以分
Fc和Fab两个部分。Fab是和抗原结合的部分,Fc是constant region,不和抗原结合,但
是和受体receptor结合。做药的时候,如果是用于neutralizing antibody,不用和受体
结合,同时由于一般antibody的分子实在太大,表达蛋白的时候就只表达一部分,也就
是fab的部分,称为fab fragment. |
|
E*V 发帖数: 200 | 44 Newly revealed Coptic fragment has Jesus making reference to 'my wife'
By Eric Marrapodi, CNN Belief Blog Co-Editor
(CNN) - A newly revealed, centuries-old papyrus fragment suggests that some
early Christians might have believed Jesus was married. The fragment, writte
n in Coptic, a language used by Egyptian Christians, says in part, "Jesus sa
id to them, 'My wife ..."
Harvard Divinity School Professor Karen King announced the findings of the 1
1/2- by 3-inch honey-colored fragment on Tuesday in... 阅读全帖 |
|
k****o 发帖数: 728 | 45 看你怎么看了。ETD确实是charge assumption的技术。如果想把precursor全部与ETD
reagent anion反应完,则需要比理论optimized reaction更多的时间,而这也同时带
来使First-generation fragment ions继续与reagent ion react产生internal
fragments和更多的small N- and C-term fragment ions。对于不大的peptide(比如
3000 Da以下的),延长reaction time来提高fragment intensity没什么大问题,因为
靠+1, +2 fragments你就能ID或characterize peptide了。 而对于大的peptide,这
样做会产生internal fragmentation(这些internal fragment你是看不到的,他们“
沉淀”成为MS2 background),会让比较small的terminal fragment ions增多(
signal增强),你只能得到比较好的terminal seq... 阅读全帖 |
|
f********t 发帖数: 4574 | 46 看了一下wiki上5.56的简介,它的杀伤力还真有问题
http://en.wikipedia.org/wiki/5.56x45mm_NATO#Performance
Criticism
There has been much criticism of the poor performance of the bullet on
target, especially the first-shot kill rate when the muzzle velocity of the
used firearms and the downrange bullet deceleration do not achieve the
minimally required terminal velocity at the target to cause fragmentation.[
14] This wounding problem has been cited in incidents beginning in the first
Gulf war, Somalia, and in the curre... 阅读全帖 |
|