M*****r
发帖数: 1536 | 1 informix
sybase
sqlite
also， MPP dbs: netezza/greenplume/vertica/... |
|
|
B*****g
发帖数: 34098 | 3 【 以下文字转载自 Programming 讨论区 】
发信人: lakewest (lake), 信区: Programming
标 题: Re: big data怎么搞定商业报表？
发信站: BBS 未名空间站 (Mon May 13 19:12:50 2013, 美东)
对于99%的传统非互联网公司来说, hadoop没有什么用, 哪有那么多的unstructure
data来处理, 上 mpp才是正道. hadoop现在基本上还是弯曲的internet 公司在用, 然
后就是一堆堆的start up用来骗funding. |
|
|
w*r
发帖数: 2421 | 5 option 1 . blanaced computing unit, bcu, such an immature implementation of
MPP...horrible architect with little or no best practice in terms of system
management and dw practice. it is soon going to be replaced by netezza which
IBM bought several years back
2. share memory database pure scale--- clustering with availability..
nothing new here
3. BLU.. is is not cpu parallelism.. in memory analytics.. or advanced cache
.. |
|
w*r
发帖数: 2421 | 6 要上MPP首选 Teradata，再选 Netezza, 再选 Vertica再选 Greenplum, 再选 Exadata
.... |
|
l******t
发帖数: 660 | 7 exadata不是纯mpp share nothing, 贵得一比, 价钱和teradata差不多, 不过卖的挺火
, 主要是从oracle rac升级容易, 很多以前用oracle的就直接买这个, dw appliance除
了teradata就是exadata份额多了.
pdw也不是那么差, 如果数据量不大, structure又 clean的dimensional model, 上pdw
也还行. v1 bug很多, v2要好一点
netezza相当不错.
Exadata |
|
w*r
发帖数: 2421 | 8 Teradata was born as MPP 20 years ago.. ...
嘿嘿 |
|
w*r
发帖数: 2421 | 9 这个是课堂问题，没有实际意义，实际工作中哪里来的distributed database???
federation 的也可以叫做distributed,
mpp也是distributed...
这种教材就是脑残 |
|
|
|
|
d******8
发帖数: 3017 | 13 谢了！
前几天做菜鸟才听了hadoop 的 web seminar from TDWI.
说是 next generation data warehousing technology，现在是RDBMS 的补充 |
|
l******t
发帖数: 660 | 14 紧急通知， 因为临时急事， 讲座从7月30日换到8月一日
版主麻烦置顶 |
|
|
y*****g
发帖数: 677 | 16 真是觉得以后的DBA 机会会变少，很多被DEVELOPER代替就可以，一些优越性能简单稳
定，就不需要太多的DBA了。我得做好转行的准备了。 |
|
|
|
c*****d
发帖数: 6045 | 19 点了上面的link，安装了GoToMeeting
现在状态是successfully connected
然后就等着？
打电话(647) 497-9391过去
输入了code之后也没动静
就是嘟嘟的声音 |
|
C********r
发帖数: 145 | 20 But it says waiting for the organizer ...Audio also say the same thing and
put me on hold until organizer starts meeting... |
|
l*****e
发帖数: 3343 | 21 wrong link??? July 17th |
|
c*****d
发帖数: 6045 | 22 me too，看来organizer还没有start meeting
我开始以为是我电脑的问题 |
|
y***c
发帖数: 676 | 23 也许meeting人数上限比较低，你打不进去了。
我正在听，看不到video。 |
|
c*****d
发帖数: 6045 | 24 Subject: SQL Server Service Broker
Time: 8:30 PM Central Daylight Time
Date: Wednesday, July 17, 2013
看起来日期不对，是不是西湖把link发错了？ |
|
|
|
|
|
|
w*r
发帖数: 2421 | 30 INNER JOIN <> EXIST
inner join 允许Cartesian,
exist 不允许
exist要求 in set要sort取distinct ,sort的cost就imply的
inner join不要求sort, 但是通常都会做sort merge join
exist对于stack tables (oracle/db2/sqlserver)优化的比较好， 对于MPP RDBMS的优
化不友好，
exist上要考虑null的side effect 特别是not exist
如果exist 的inset本身是unique index, 其cost和inner join 一样 |
|
d**********3
发帖数: 1186 | 31 Please contribute your 3 minutes to help us finish the session(MPP/Hadoop)
survey if you attended last night's session, Thanks
http://www.surveymonkey.com/s/RBJZRRZ |
|
d**********3
发帖数: 1186 | 32 Please contribute your 3 minutes to help us finish the session(MPP/Hadoop)
survey if you attended last night's session, Thanks
http://www.surveymonkey.com/s/RBJZRRZ |
|
y***c
发帖数: 676 | 33 Done.Thanks for your good job! |
|
c*****d
发帖数: 6045 | 34 Done, thanks for the seminar |
|
|
c*****d
发帖数: 6045 | 36 不知道，你问问西湖
我进去的时候都讲了30分钟了 |
|
w*r
发帖数: 2421 | 37 Is there a recording of the session? I missed the meeting . Appreciate if
someone can provide a recording. Also as many of you may know I am experts
in mpp tear data . Currently as lead dw architect. If someone one this board
who is interested to be junior data analyst I would like to know you and
may refer you in my group. |
|
w*r
发帖数: 2421 | 38 Most rdbms like oracle db2 SQL server optimizer is designed in the way to
shorten the io path cost to accommodate oltp. oracle index join hint will
bring both index into consideration. Bitmap index provided a faster
operation using and/or operation on the single index, however I doubt it
will benefit index join performance . Bitmap index probably more cost
effective when the distinct values in the index is low.
For low cardinality columns , partition is a better choice comparing to
index in mos... 阅读全帖 |
|
w****w
发帖数: 521 | 39 解还是有的，就看你肯花多大本钱。如果xxx经常重复搜索，可以建一个index table。
也可以按列中的词做索引。最终的办法是搬到hadoop或greenplum那样的MPP系统去。 |
|
w****w
发帖数: 521 | 40 解还是有的，就看你肯花多大本钱。如果xxx经常重复搜索，可以建一个index table。
也可以按列中的词做索引。最终的办法是搬到hadoop或greenplum那样的MPP系统去。 |
|
w****w
发帖数: 521 | 41 解还是有的，就看你肯花多大本钱。如果xxx经常重复搜索，可以建一个index table。
也可以按列中的词做索引。最终的办法是搬到hadoop或greenplum那样的MPP系统去。 |
|
w****w
发帖数: 521 | 42 解还是有的，就看你肯花多大本钱。如果xxx经常重复搜索，可以建一个index table。
也可以按列中的词做索引。最终的办法是搬到hadoop或greenplum那样的MPP系统去。 |
|
m*******9
发帖数: 103 | 43 组里急需DE contractor，收到几十份agency提交的简历，清一色的烙印。来这里看看
有没有朋友感兴趣？学校不需要很好（任何一个美国学校都秒杀绝大多数印度学校），
专业无需CS，但要有几年相关经验。 由于工作强度大， 一定要有兴趣。very
competitive compensation. Core skills: SQL and Python. 有意或有问题发邮件到
： [email protected]/* */
*******
Build data expertise and own data quality for the awesome pipelines you
build
Architect, build and launch new data models that provide intuitive analytics
to your customers
Design, build and launch extremely efficient & reliable data pipelines to
move data (both large ... 阅读全帖 |
|
|
b******n
发帖数: 4509 | 45 这么算下来，直接 mpp + dell advantage 最后含税价只有 $670 超值啊
话？ |
|
a***e
发帖数: 27968 | 46 经过一番chat,搞定了，把$600的gift card花出去了
这个东西是这个样子的
通过MPP/EPP买，减14刀，所以起价685.99,$50off争取一把，今天有新coupon,
虽然网上不能用在XPS8700.正好抵税,最后价格685
加上免费的Dell Advantage 2nd free shipping和5%back
到时候回扣$30
还不错 |
|
|
mn
发帖数: 46 | 48 If you're interested in working for SAS, #1 best place to work. PM me.
Thanks.
===
Job Description
Software Developer-11003071
Description
Join the world's leading statistical software company and make a difference
in the way that analytical software is developed and deployed. SAS is
expanding its software for high-performance analytics.
In this position you will develop and support SAS software components for
high-performance computing. You will work closely with Ph.D. specialists in
statistics... 阅读全帖 |
|
c****e
发帖数: 1453 | 49 Taobao used Greenplum's product to replace Oracle because Oracle is too slow
. They claimed a used-to-be 2 hour BI query takes only minutes to finish now
. The current trend is big data analysis, where you get MPP architecture and
in-memory DB.
Non-SQL approach wins in many cases but also has its limitation. The
consistency and transaction support might be very expensive. That's why you
see more and more solutions in BI area are moving to Non-SQL but most of the
billing pipelines are still on re... 阅读全帖 |
|
