发信人: a060601199 (昵称), 信区: JobHunting
标 题: google, vmware 面经
发信站: BBS 未名空间站 (Wed Nov 27 03:43:34 2013, 美东)
google面的是SRE
电面是国人大哥,一些c语言的pointer问题,然后一道leetcode原题
onsite1:
1)combination,还是没dupliate的,要把结果保存起来,我说用linked list,因为是
用c写,还要自己implement linkedlist, 略坑爹
2) 就是很简单的统计两个string分别有多少个单独的letter
onsite2:
bst inorder iterator
onsite3:
一个文件,每行是rack_name + machine id,输出每个rack有多少个machine,按大小排
序,我是先扫一遍存hashtable,再存进linkedlist再sort,这回没让我实现hashtable
跟linkedlist了,不过要我把用过的api单独再declear一下,最后再写个mergesort
onsite4:
有很多个machine,要求检测哪些die了,要求parallel,就写了一个for loop创建若干
个thread来执行任务,有点thread pool的感觉,用一个array来表示哪个machine被检测
了与否,最后再写个检测的函数,要求socket programming,我就用tcp connect了一下
,里面的各种api(pthread_thread, connect)我都不记得了,随便写写意思意思
onsite5,system design
1) 电话号码查询,其实就是hashtable,先一个machine再scale,consistent hash跟
redundancy
2)如何提高disk 数据访问的latency,我给的答案有 1,cache, 2 seqencial visit 3
, redundancy disks一起访问,用最快返回的数据
3)连不上server了怎么trouble shooting,这题坑太大了,就随便答答
HR的反馈是我虽然代码写的快但考虑的不够仔细,我估计是第四轮吧,呵呵
vmware面的是 core storage
电面1:
前面问了些scheduler,mutex,tcp windows的问题,code题有3个
1) memcopy,最简单的,不用考虑overlap
2) 如何判断stack是往上还是往下增长,答案是call一个子函数,返回子函数的local
variable的address,然后跟母函数的local variable address比较
3) bubble sort
电面2:
描述syscall怎么实现,argument存哪,在kernel里面怎么访问user land的地址
device怎么与os communicate
coding 题
1)把{0,1,0,1,1}变成{0,0,1,1,1},这叫partition还是叫啥,好像leetcode有
2)Flatten Binary Tree to Linked List
onsite1:
how to implement/walk through the process of syscall, signal, read a mounted
file等等
onsite2:
1)
detect loop in linked list
give an array contain number from 1-10000, but just one num is missing, find
the number
answer: add the sum
2)
c struct padding/size, which the flag in compiler for optimization padding?
fpack-struct
3)
detect loop in linkedlist
onsite3:
1)
implement barrier by mutex and semaphore(wake up one by one)
我大概是这么写的;
mutex.lock();
count++;
if (count == N) {
mutex.unlock();
} else {
mutex.unlock();
semaphore.wait()
}
semaphore.signal();
2)
implement cognitional variable by semaphore and mutex,我不记得我怎么写了
onsite4:
写个malloc/free/calloc。。。
how to sync caches in two machines, cache size is limited, said, machineA
want to write to cache, but maybe machineB don’t have enough cache. answer:
need to flush cache in machineB(大概是这么个意思,没时间了,没仔细问)
onsite5:
spin lock fair solution? answer: use bitset to represent which thread
already held the lock before
sync with two machine’s disk?
answer: write ahead log first, if another machine also succeeded write to
log then write data to itself disks and clear log
--
※ 来源:·WWW 未名空间站 海外: mitbbs.com 中国: mitbbs.cn·[FROM: 148.]
Wednesday, November 27, 2013
google, vmware 面经
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment