跟讀練習: Databricks Interview Questions Part 1 - 通過影片學習英語口語

選擇練習模式

每種模式都會開啟全屏練習室。

186 句

如果句子過短或過長，請點選 Edit 進行調整。

hello and welcome to everybody on cloud fitness

so this video is a very on demand video where i

am starting a play series on the data factory interview questions

so a lot of you have sent in emails to me right

you know to make videos on azure data factory

and databricks interview questions so

that is why i have already created a video of a playlist for azure data factory

and this one is for the databricks so let's move ahead

and see what are the questions

that you know we we can start off from this particular video right

so the very first basic question that I have is what is Databricks

so this question might be asked in the interviews just to understand how much you know Databricks

so you should be able to answer it really well right

so when we talk about is your Databricks it is an Apache Spark based environment

that is what the like the very first thing is

that you should know that it is an Apache Spark based environment it is a collaborative environment where when I say collaborative,

it essentially means that you have your data engineers, you have your data analysts,

you have your data science people coming in all together at the one place and working on the data, right?

It is an Apache Spark based big data analytics service.

When I say

Apache Spark, whatever code that you write in Databricks, that is basically run on Spark, right?

It is a Spark code that you write.

Now, apart from that, it is a big data analytics service it's not just the structured data

that you can process it's you can process semi -structured

and even unstructured data as well in the databricks now the second thing is um you know that

it supports multiple language

so it is not very much language bound you can work on scala

you can work on r you can work on python you can work on sql as well

so these are the things

that you should definitely know about databricks as well as it

can you know easily integrate with your other Azure services like

your you know your active directory your keywords your storage accounts

if you have data in your storage accounts you can directly go ahead link it

and use that data right so it is a very essential

and useful SaaS offering so in fact let us go to the second question

that is also very commonly asked because

when you are writing Databricks in your resume what usually happens

is a person who has worked on Databricks essentially would know about Data Factory as well

So this is a very common question that what is the difference between Databricks and DataFactory?

Because whatever code you write in Databricks, let's say you're trying to do a filter,

group buy or anything, that can essentially be done in DataFactory, right?

Or vice versa.

So why do you need Databricks when it can be done in DataFactory?

So you should be very clear.

It is essentially a very, you know, both are actually different offerings.

The very first difference is the purpose, right? the data bricks has a different purpose

and data factory has a different purpose data factory is essentially

for orchestration for the data flow for the data transformation for the code free data transformation

and orchestration but

when you talk about data bricks i have already explained

it is an collaborative platform where multiple people can come in together data science team can come

and they can train their models they can train their models in data bricks

when you talk about data engineers they can write a code solution

so code solution is actually not bad

because you have a lot of flexibility uh you know with

when you write your own code which you are

which you don't get with the data factory right so

that is essentially one of the reason that most of the organization

when there is any kind of you know data engineering task

or data engineering transformations

that have to happen on the code

they prefer using data bricks uh rather than data factory data factory is mostly

you know most of the organization they use for uh data movement

and orchestration purpose so this is the very key and major difference

when you talk about data breaks and the data factory

and of course

when you talk about data processing as well right in data uh breaks

you can process not just your structured semi -structured

or even your you know your batch data you can easily uh

you know use your streaming data as well though in data factory also

you have an option uh to use streaming data

but live streaming is possible only in databricks as of now right

so even when you talk about data processing databricks has an advantage

but yeah these are the two uh major things apart from

that there are cosmetic things that you can talk about like you can say

that data factory uh you know is a code free etl solution

and those kind of things but these two are the major major differences

that you should definitely cover

and then a very you know common question is describe spark architecture

so if you're working on databricks you should know spark architecture right

so i have already in fact explained about the spark architecture in one of my videos let me check that out

So if you see this runtime architecture of Spark Databricks, I have already explained, you know, in detail regarding the architecture of Spark.

If you want to know details on this, you can go ahead and you can, you know, just watch out this video as well.

In this, I have, you know, explained everything about the runtime architecture of Spark, how everything works and how you're.

In fact, I created this diagram to show, you know, the flow of the basically flow of,

you know, whenever you write a query, what is the flow that happens in your data breaks?

So going back to the PPT, if you see over here, right, when I talk about a Spark architecture,

what essentially happens is you write a code.

I'm actually trying to explain this in very short

if you want to have a very detailed explanation you can watch

that video so

when you talk about spark architecture you write your code whenever

you write your code you write it through spark context you uh you know you write it on spark context

and with the help of driver program driver program actually

or uh you know you you write it through spark context

you know whenever you must be writing code in data breaks you you would write

that spark .rd right sc .parallize so SE is basically spark context right

so you create a driver program

and then behind the scenes what happens is spark driver submits the job submits your code

100

you know submits your code to dark scheduler so

101

when I say dark scheduler dark scheduler is nothing

102

but directed acyclic graph scheduler

103

so what it what it does is it essentially takes your code what you have written

104

and it tries to create a logical and physical plan out of it

105

when i say logical and physical plan out of it it tries to optimize your code

106

and tries to create a plan out of it on how it will execute your code

107

now once it has created your physical and the logical plan then what happens it it says the

108

you know it breaks your code in form of tasks

109

and your stages right it breaks your code in form of stages

110

and tasks and these tasks are actually sent to your executors

111

or to your worker nodes for the execution purpose right

112

and then from there executors actually write your data to whichever location you have specified

113

and in between cluster manager actually manages your you know all the worker nodes

114

and the executors right

115

so this is a you know very uh short introduction of how spark architectures behave

116

you have your code you write it through you know spark context

117

you have your driver program you send your code uh

118

you know then you have your cluster manager cluster manager tries to you know make sure

119

that you have worker nodes ready

120

and like it manages your worker nodes basically worker nodes are nothing

121

but a set of machines where your actual code will execute

122

now your code is there with the help of spark driver it is sent to

123

that scheduler which creates logical

124

and physical plan out of it once it is done

125

it actually you know segregates your code in form of stages and tasks

126

which are supposed to be executed on the executors

127

then all those tasks through the with the help of task

128

scheduler are sent to your executor nodes for the execution purpose

129

so this is basically a very short introduction of spark architecture

130

and then coming on to the next question is what is the maximum size of worker node that you have used

131

Now, this is a very, very, very important question.

132

In fact, the most important one, I would say.

133

Now, this essentially shows everything, you know, about how much you have worked on Databricks and how much you know about Spark.

134

Essentially, this question answers everything.

135

So, when I ask or when anybody asks you what is the maximum size of worker node you have used,

136

it shows how much data you have been using it shows

137

you know how you how good you are in configuring the cluster

138

it shows how much you understand the spark architecture right so whenever this question is asked you should not

139

exceed the worker node configuration more than 64 gb that is something

140

that you should keep in your mind what is the maximum size of worker node

141

you have used essentially it depends on your project as well

142

but most of the project they do not need more than 64 gb

143

or worker nodes right because what happens is

144

if you increase the size of your worker nodes more than 64 gb there is a garbage garbage collection overhead

145

which reduces your performance and mostly you will not need

146

that right you will not need more than 64 gb after

147

this you can you know go ahead increase the number of worker nodes

148

or you know you can increase the number of cores you have

149

and there are a lot other optimizations to be done

150

but 64 is a cross checkpoint like more than 64 gb is actually not required

151

and in case if you want to know more more about this why in more than 64

152

and all for that also i've created you know, one video on how to select worker and driver type in Databricks.

153

So if you go through this particular video, you will understand end -to -end how to choose a cluster, basically.

154

How you can go ahead, choose a cluster, and what cluster configuration best suits for each workload.

155

But yeah, a straightaway answer is 64 GB.

156

If you want answer, detailed answer of why, you can go ahead and check the video out.

157

And then you have the next question is, how do you choose cluster configuration?

158

So when I say how do you choose cluster configuration, again, this is a very important question and it has,

159

you know, you can speak about it for almost half an hour.

160

But yeah, again, for choosing the cluster configuration,

161

you should understand that whenever you are trying to do any normal ETL or ELD work,

162

in that case, you should go for memory optimized clusters right after

163

that whenever you have some development tasks that you need to do

164

and quickly check it out in

165

that case always go for general purpose clusters then the third thing is

166

if you have a strong ETL task or ELT task

167

which requires a lot of shuffle it has a lot of joins in

168

that case always go for storage optimized cluster

169

so these things you should always remember these are the key points actually in fact i can actually go

170

and show you so whenever you are in the databricks portal

171

and you try to create a cluster

172

so let me try to create a cluster and in fact show you

173

if you go to the worker type

174

or the driver type you can actually see there is general purpose

175

there is general purpose hdd hdd you should never choose never choose an hdd option

176

then you have this memory optimize over here and then after you have storage optimized right

177

So these things you should always keep in mind that before attending any interview for data breaks.

178

So this really makes sense that for any normal EGL workload, you should always use memory optimized,

179

storage optimized for a shuffle intensive workload because it has a disk attached to it.

180

So it can easily speed up your process of any shuffle that is required.

181

That is why you should go for a storage optimized once in case of shuffle intensive work.

182

So, this is pretty much that I wanted to cover in this particular video.

183

And thank you so much for being till here.

184

Do let me know in the comment section if you have any issues, any doubts.

185

And do remember to like, subscribe and share this particular channel.

186

Thank you so much.

下載應用

流利口語所需的一切

AI 發音評分逐句打分

音標練習掌握每個音

詞彙學習建立你的詞庫

詞彙遊戲邊玩邊學

掃碼

Download on theApp Store Get it onGoogle Play

FOR YOU

單詞	發音	釋義
spark 名詞	/ˈspɑɹk/	火花
cluster 名詞	/ˈklʌstɚ/	串, 叢 /叢
worker 名詞	/ˈwɝ.kɚ/	工人, 工作者
node 名詞	/noʊd/	結節 /結節
architecture 名詞	/ˈɑː.kɪˌtɛk.t͡ʃə/	建築 /建築, 建築學 /建築學
essentially 副詞	/ɪˈsɛnʃəli/	本質上 /本質上, 基本上
configuration 名詞	/kənˌfɪɡ.əˈɹeɪ̯.ʃən/	編排 /編排, 配置
brick 名詞	/ˈbɹɪk/	磚仔 /磚仔, 磚 /磚
submit 動詞	/səbˈmɪt/	讓步 /讓步, 屈服
transformation 名詞	/ˌtɹæns.fɔɹˈmeɪ.ʃən/	轉型 /轉型, 轉變 /轉變
execute 動詞	/ˈɛksɪˌkjuːt/	處決 /處決, 處死 /處死
collaborative 形容詞	/kəˈlæb.ə.ɹə.tɪv/	協作的 /協作的
execution 名詞	/ˌɛk.sɪˈkjuː.ʃən/	執行 /執行, 實行 /實行
garbage 名詞	/ˈɡɑɹ.bɪd͡ʒ/	垃圾, 廢物 /廢物
analytics 名詞	/ˌæn.əˈlɪt.ɪks/	分析學 /分析學

影片中出現的短語動詞

單詞	發音	釋義
go ahead 動詞	/ˌɡoʊ əˈhɛd/	請便 /請便
come on to 動詞		輕薄, 吃豆腐
go through 動詞		通過 /通過
speed up 動詞	/spiːdˈʌp/	加速, 加快
watch out		小心！

影片中的語法

說話人最常用的結構，並附上影片中的原話：

結構	影片中的用法
現在完成時 have/has + 過去分詞 — 過去發生但與現在仍有關聯的事	have sent · have already created · has worked
被動語態 be + 過去分詞 — 強調發生了什麼，而不是誰做的	might be asked · be done · can be done

需要注意的發音

說話人用了 6 次縮略和弱讀形式，例如 you're, don't, I'm。請按聽到的簡短形式來說。

“sh” 和 “zh” 音: essentially /ɪˈsɛnʃəli/, configuration /kənˌfɪɡ.əˈɹeɪ̯.ʃən/, transformation /ˌtɹæns.fɔɹˈmeɪ.ʃən/, execution /ˌɛk.sɪˈkjuː.ʃən/, explanation /ˌɛkspləˈneɪʃən/
長單詞——注意重音位置: architecture /ˈɑː.kɪˌtɛk.t͡ʃə/, essentially /ɪˈsɛnʃəli/, configuration /kənˌfɪɡ.əˈɹeɪ̯.ʃən/, transformation /ˌtɹæns.fɔɹˈmeɪ.ʃən/, collaborative /kəˈlæb.ə.ɹə.tɪv/

中文母語者容易讀錯的音：

/v/ — 不要讀成 /w/，上齒輕觸下唇: collaborative /kəˈlæb.ə.ɹə.tɪv/, intensive /ɪnˈtɛnsɪv/, behave /bɪˈheɪv/, overhead /ˈəʊvəˌhɛd/, whichever /wɪt͡ʃˈɛvɚ/
詞尾子音 — 要讀清楚，後面不要加母音: spark /ˈspɑɹk/, node /noʊd/, brick /ˈbɹɪk/, submit /səbˈmɪt/, execute /ˈɛksɪˌkjuːt/
/r/ 和 /l/ 的區別: collaborative /kəˈlæb.ə.ɹə.tɪv/, portal /ˈpɔːɹtl̩/

如何用這段影片練習

先完整聽一遍影片，不要開口，記下不認識的單詞。
先用 0.75 倍速逐句跟讀，熟練之後再回到正常速度。
錄下自己的聲音並與原聲對比，特別注意 spark, cluster, worker 這類單詞。

什麼是跟讀法？

跟讀法 (Shadowing) 是一種有科學依據的語言學習技巧，最初開發用於專業口譯員的培訓，並由多語言者Alexander Arguelles博士普及。這個方法簡單而強大：您在聽英語母語原聲的同時立即大聲重複——就像是一個延遲1-2秒緊跟說話者的影子。與被動聽力或語法練習不同，跟讀法強迫您的大腦和口腔肌肉同時處理並模模擬實的講話模式。研究表明它能顯著提高發音準確性，語調，節奏，連讀，聽力理解和口語流利度——使其成為雅思口語備考和真實英語交流最有效的方法之一。

影子跟讀法: 閱讀完整分步指南 →

跟讀練習: Databricks Interview Questions Part 1 - 通過影片學習英語口語

選擇練習模式

下載應用

推薦影片

Excellent 2 Years AWS DevOps Engineer Mock Interview with Questions and Answers | Cloud | Docker

Databricks Interview Questions by Sr Data Architect

Uber Data Engineering Mock Interview - Ride-Sharing Data Warehouse Schema

FULL Data Science Mock Interview | SQL, Statistics, Business

Distributed Task Scheduler: System Design Interview (Stripe & Amazon Offers)

.NET Mock Interview - Scenario Based - 7+ Years Experienced Candidate

Top 30 IICS interview questions and answers | Informatica cloud interview questions full video #iics

Learn how to SOLVE a data analytics case study problem

Google product manager behavioral interview (Top 5 questions)

10 Data Analyst Interview Questions and Answers - Senior Data Analyst Explains

🔥 Java Multithreading Rapid Fire | Can You Answer These Interview Questions? #shorts

Incident Management Process: A Step by Step guide

本課的詞彙與口語要點

影片中的重點詞彙

影片中出現的短語動詞

影片中的語法

需要注意的發音

如何用這段影片練習

什麼是跟讀法？