Practice Free H13-711_V3.0 HCIA-Big Data V3.0 Exam Questions Answers With Explanation

group by

coalosce

onnect

join

1

2

3

4

HDFS fsck

HDFS fsck-delete

HDFS dfsadmin -report

HDES balancer -threshold 1

Unified Authentication Server

Unified authentication management module

Identity information storage server

Unified Session Management Module

Install Manager->Execute preinstall->LLD tool for configuration->Install cluster->Check after installation->Configure after installation

LLD tool to configure -> execute preinstall-> install Managers-> install cluster-> post-installation check-> post-installation configuration

Install Manager->LLD tool for configuration->execute preinstall->install cluster->post-installation check->post-installation configuration

LLD tool to configure -> execute preinstalls-> install cluster-> install Manager-> post-installation check-> post-installation configuration

The secondary index associates the column to be searched with the rowkey into an index table

At this point, it is listed as a new rowkey; the original rowkey becomes the value

The secondary index is queried twice

all of the above

Datastream API, DataSet API

Data batch API.DataStream API

Stream API.Batch API

Batch API, Stream API

Fusioninsight Miner

Fusioninsight Farmer

Fusioninsight HD

GaussDB 200

Once the table is created, the table name cannot be changed.

Once the table is built. No new columns can be added

The external keyword needs to be specified when creating an external table

Once the table is created, the column names cannot be changed

Hive

YARN

HDFS

Zookeeper

HMaster

Slave

HRegionServer

Data Node

digital government

Smart Park

smart transportation

finance

Spark SQL is compatible with most Hive syntax and functions

Spark SQL cannot use Hive's custom functions

The execution engine of Spark SQL is Spark core, HiveThe default execution engine is MapReduce

Spark SQL relies on Hive metadata

HDFS

HBase

Spark

YARN

HDFS

NameNode

DataNode

Hbase

HDFS

NameNode

DataNode

HBase

When creating TopicE of Kafkal, the number of Partitions must be set

When creating TopicE of Kafkal, the number of Partition copies must be set

Setting up multiple replicas can enhance the disaster tolerance of Kafka services

all of the aboveA. True

Dog 1, owl 1

Cat 2, dog 4, owl 2

Cat 2, dog 3, owl 1

Cat 1, cat 1, dog 2, dog 2, owl 2

Avro sink

Thrift sink

Hive sink

Null sink

HLog

MemStore

HFile

HMaster

Kafka cross-cluster data synchronization scheme

Kafka data backup within a single cluster

Kafkat but intra-cluster data recovery

None of the aboveA. True

graph computation

Interactive query

batch processing

real-time stream processing

UNION ALL

GROUP BY

SELECT

JOIN

Data can put different data into different buckets according to the way of buckets

Unsortable in bucket

The advantage of bucketing is that it can achieve higher query processing efficiency and make sampling more efficient

You can specify the number of buckets when creating a table

DataNode

ZooKeeper

HMaster

HRegionServer

Multi-node deployment

Statistical analysis and data mining

Security Control Based on Kerberosi Certificate

Offline and real-time data processing

scard

sunion

zcount

hexists

Active NameNode

Standby NameNode

DataNode

All of the above are wrong

Memory Channel

JDBC Channel

File Channel

HDES Channel

business model discovery

value for evaluation

Products Featured

Operational Analysis

True

False

True

False

High-performance computing, ARM-compatible high-performance Kunpeng processors and x86-architecture servers and solutions

Safe and reliable, creating high quality as stable as Mount Tai

Open ecology, support mainstream software and hardware in the industry, and work with developers, partners and industrial organizations to create a new base for intelligent computing

everything abovecorrect

HTML

XML

two-dimensional table

JSON

A DataSet is a strongly typed collection of domain-specific objects

DataSet can perform most operations without deserialization

DataSet needs to be deserialized to perform operations such as sort, filter, shuffle, etC.

DataSet - highly similar to RDD. Better performance than RDD

system security

Authority authentication

Data Security

everything aboveA. True

Kafka cross-cluster data synchronization method

Kafka data backup within a single cluster

Kafka data recovery within a single cluster

None of the aboveA. True

spark streaming

spark core

mapreduce

spark sql

Hive finally stores data in HDFS

Hive is a data warehouse tool for the Hadoop platform

HQL can execute tasks through MapReduce

Hive has a strong dependence on HBase

reduceByKey

filter

map

saveAsTextFile

Real-time online data analysis

Non-real-time analysis, such as log analysis, statistical analysis

Data mining, such as user behavior analysis, interest division, regional display

Data aggregation, such as daily, weekly user clicks,Click to rank

When creating a topic, the number of replicas must not be greater than the number of currently surviving Broker instances, otherwise the topic creation will fail

When the Producer of Kafkal sends a message, it can specify which Consumer consumes the message

Kafka will store metadata information in Zookeeper for

After Kafka is installed, the sensitive data storage directory cannot be configured.

Distributed memory computing engine

Distributed file system

Unified scheduling of cluster resources

Stream processing capabilities

Hot Standby for Primary NameNodel

Standby NameNode) has no memory requirements

Help the main NameNodet merge edit logs and reduce the startup time of the main NameNodel

The standby NameNode should be deployed to the same node as the primary NameNode

Extensibility

High reliability

high performance

Ease of use

Follower

Controller

Brocker

Leader

After obtaining this input stream, the client calls the read method to read the data. The input stream selects the nearest DataNode to establish a connection and read data.

The client calls close. to close the input stream.

The location where the data block corresponding to this file in NameNodel is obtained by calling NameNode remotely through RPC

If the end of the data block has been reached. Then close the connection with this DataNodel, and then re-find the next data block. until all data is read.

The client calls the open method of the FileSystem instance to obtain the input stream corresponding to the file.

Lots of small file storage

High capacity and high leaf swallowing capacity

low latency read

Streaming data access

create table

Import Data

subquery

JOIN Cha Xun

High cost of data storage

Insufficient streaming data processing performance

Limited scalability

Batch data processing is missing

HBasePeriod refreshMemstore, silentthink period is 1Hour

When the number of files in WALs reaches a threshold

The total size of the MemStore in the Region has reached the preset Flush Size threshold

The ratio of the total memory occupied by the MemStore to the total memory of the RegionServer exceeds the preset threshold size

After obtaining this input stream, the client calls the read method to read the data. The input stream selects the nearest DataNode to establish a connection and read data.

The client calls close. to close the input stream.

The location where the data block corresponding to this file in NameNodel is obtained by calling NameNode remotely through RPC

If the end of the data block has been reached. Then close the connection with this DataNodel, and then re-find E. a data block. until all data is read.

The client calls the open method of the FileSystem instance to obtain the input stream corresponding to the file.

TRUE

FALSE

True

False

True

False

content

document

piece

None of the above

Farmer is a data service framework

GaussDB is an open source database product

Fusioninsight Manager is a distributed system management framework, administrators can control distributed clusters through multiple access points

Fusioninsight HD is an enhanced version based on the open source big data software Hadoopl

Data is scattered among various departments of the enterprise, and the same data is stored in different formats within each department.

Diversified data structures.

technological advancements of competitors.

Scattered data has problems such as noise, missing, and non-standard storage types, which requires a lot of data preprocessing.

DateNode

Nodemanager

Driver

ResourceManager

It is impossible for each node of a kafka cluster to be l with each othereader and flower

If the leader fails, other followers will take over (become the new leader)

Because the leader server carries all the request pressure. Therefore, from the overall consideration of the cluster, kafka will distribute the leader evenly on each instance to ensure the overall performance is stable

Kafka needs to select a leader for partition replication, and the leader is responsible for reading and writing partitionsD. operation, other replica nodes are only responsible for data synchronization

True

False

Socket streams

JDBC

Files

Ccollections

True

False

True

False

True

False

Supervisor is the component that receives data in topology and then performs processing

There is a logic error in the topology business, and it cannot run normally after submission

The topology is too complex or the number of concurrent users is too large, resulting in workerThe startup time is too long, exceeding the waiting time of Supervisort

The supervisor's slots resources are exhausted, and after the topology is submitted, the slots cannot be allocated to start the worker process.

hadoop

head

bigdesk

IKAnalyzer

Operating system security hardening

Component data encryption

data integrity check

User authority authentication management

True

False

True

False

True

False

True

False

Shared file system

Object Storage

HDFS

Local file system

True

False

True

False

True

False

True

False

RDD is a read-only, partitionable distributed dataset

RDD is Spark's abstraction of underlying data

RDDs have a lineage mechanism (Lineed)

RDDs are stored on disk by default

TRUE

FALSE

Data Lake Governance Center DGC

Cloud operating system FusionSphere Openstack

AI development platform ModelArts

Video Analysis VAS

True

False

Supports threshold control for the total number of established connections

Supports threshold control for the number of connections established by each user

Supports threshold control on the number of connections established by a specific user

Supports threshold control of the number of connections established per unit time

sem1-persistent

ephemeral

temporary

persistent

Key

Value

Timestamp

KeyType

Visual modeling helps DBAs in modeling design and lowers the threshold for using HBase

Qualifier HBase column: each column represents an attribute of business data

Realize the division of labor: DBAs focus on data table modeling, developers focus on user table names and columns used

Column user table column: each column represents a KeyValue

The port number

IP address

server nickname

username

HFile

TextFile

SequenceFile

RCFile

Flink

HBase

Hive

HDFS

FIFO scheduler

capacity scheduler

Fair scheduler none of the

At Most Three Times

Exactly Once

At least once (At Lease once)

At Most once

cArbon is also a high-performance analytics engine that integrates data sources with spark.

cArbon uses a combination of lightweight compression and heavyweight compression to compress data, which can reduce data storage space by 60%-80% and greatly save hardware storage costs.

arbon is a new Apache Hadoop native file format that uses advanced columnar storage, indexing, compression, and encoding techniques to improve computational efficiency to help accelerate data queries over petabytes of magnitude, and can be used for faster interactive queries.

The purpose of using carbon is to provide ultra-fast responses to ad-hoc queries on big data.

Reduced key-values

Turn off persistence

Limit Redis memory size

slowlog configuration

Region load balancing, Region splitting and Region allocation after splitting

Responsible for creating tables, modifying tables, deleting tables

Responsible for load balancing of RegionServer

Region after RegionServer failsmigrate

Support GDS tools to speed up data storage

Ensure high reliability of data and systems

Trillions of data correlation analysis seconds response

Unified management console

data collection

data storage

data management

data analysis

data presentation

Many data sources and formats

Fast data growth and fast processing

Large amount of data, large amount of calculation

Low data value density, high value

Complete the hardware installation

Complete the node host OS installation

Prepare tools and software. Such as Putty, LLD, Fusioninsight HD software installation package, etC.

Prepare planning data. such as network parameters and role deployment locations

Stream API, Batch API

Data Stream APl. DataSet AP

DataBatch AP1.DataStreamAPIi

BatchAP1, Stream APi

Check the client's exception stack to determine whether the client is using problems

Check the running log of the main Nimbus to determine whether the Nimbus server is abnormal

Check the Supervisori running log to determine whether the Supervisor is abnormal

View Worker running log

yarn.nodemanager.local-dirs

yarn.nodemanager.log-dirs

yarn.acl.enable

yarn.log-aggregation-enable

Miner

Spark

Hive

Flink

Each Executor (thread) can run a number of tasks (tasks)

Each Executor (thread) can run different components (spout or bolt) of the task (task)

Each worker can run multiple Executors (threads)

Each worker can only run Executors (threads) for one topology

partition

sort/merge

????

ccombine

True

False

True

False

string

Sat

Hash

sorted set

True

False

Postingonesubscriptioninformationsystem, the sameNumber of barsData can be consumed by multiple consumers. data isConsumptionnot laterdelete immediately

Distributed Messaginghand overThere are two mainwantofmessage passing pattern,peer to peertransferpattern, haircloth-subscription model

point-to-pointinformation systemmedium, cancanhave multiple consumptionsat the same timeremovefeedata, becauseThis does not guarantee the order in which data is processed.

In a point-to-point messaging system, when a messagefeeByremovefeeteamone of the columnsdataAfter that, thedata rulefromdelete from message queue

full-text search

Term-based search

Search based on score

Retrieval based on metadata

True

False

True

False

True

False

Saccn? ?node periodically fromMa? ?download?and?files

in HDFS Sectionformat onceRear,?will generate fs?me and?twodelivery

to replace?file, and will?coincident name?make up new?and and.

to avoid?Increasing,?will be merged periodically?become new?.

put frimage?combine into new? ?document.

Import Data

export data

volume reduction

Expansion

True

False

lower diskI0

Enhanced security

to mentionHigh reliability

lower gridI0

supportTez, Spark and other computing engines

Can query and manage petabytes of distributed data

E for dataTL process automation

Direct access to HDFS files and HBase

True

False

True

False

True

False

True

False

True

False

True

False

True

False

log.retention.bytes

server.properties

log.cleanup.policy

log.retention.hours

True

False

True

False

TRUE

FALSE

to enterA series of operations such as line cleaning, word segmentation, and establishment of an inverted index

Provides the ability to search full text, conditions can include words or phrases

The written data can be checked in real timeSow

numberOptional rewrite when data is writtendeleteand compression function

True

False

Modulo conversion

Null conversion

splice conversion

Add constant field

True

False

True

False

True

False

to foldGeneration computing

stream processing

datastorage

batch processing

True

False

True

False

True

False

True

False

Updates sent by the client are applied in the order in which they were sent

A message is to be received by more than half-respected servers,he will be able to successfully write to disk

A message update can only succeed or fail. There is no intermediate state

The number of Zookeeper nodes must be odd

Loader service master node

The node on which the YARN service job is running

Both of the first two are required

Neither of the first two are needed

The number of Zookeeper nodes must be odd.

Updates sent by the client are applied in the order in which they were sent.

Message updates can only succeed or fail, with no intermediate states.

A message needs to be received by more than half of the servers.it will be able to successfully write to disk

When creating an external table, you need to specify the external keyword

Once the table is created, the table name cannot be changed

Once the table is created, the column names cannot be changed

Once the table is created, no new columns can be added

During Spliti

During Flush

ompactionj process

HFile separation process

CAS

AOS

Kerberos

LDAP

Assign roles to users

Configure permissions for roles

Assign roles to user groups

Configure permissions for user groups

3

4

6

none of the above

Iterative calculation

Offline computing

real-time interactive computing

Streaming Computing

Flexible and convenient ETL

Only supports MapReduce computing engine

Direct access to HDFS files and HBase

Easy to use and easy to program

Zookeeper

HDFS

Spark

HBase

Non-real-time analysis. Such as log analysis, statistical analysis

data mining. Such as user behavior analysis, interest analysis, regional display

Data summary. such as clicks per user per day,Click to rank

Real-time online data analysis

Configure how jobs connect to external data sources

Configure how jobs connect to internal data sources

Provide optimization parameters to improve data import and export performance

Make sure there are conversion steps

Barriers are periodically inserted into the data flow and flow with it as part of the data flow

Barriers are at the heart of Flink snapshots

A barrier separates the snapshot data of the current cycle from the snapshot data of the next cycle

When the barrier is inserted, it will temporarily block the data flow

5

6

3

4

sink

topic

Source

Channel

As long as one HiveServer is unavailable, the entire HiveEcluster is unavailable

HiveServert is responsible for accepting client requests, parsing, executing HQL commands and returning query results

MetaStore is used to provide raw data services and depends onDBServer

At the same time, only one HiveServer is in Active state, and the other is in Standby state

You can specify a specific user for log collection, for example, only download logs generated by UserA.

You can specify a time period for log collection, for example, only collect logs from 2016-1-1 to 2016-1-10.

You can specify an instance for log collection, for example, specify to collect metstore logs.

The node IP can be specified for log collection, for example, only the logs of a certain IP can be downloaded.

sequence_file

text_file

Cbinary_file

graph_file

Service level configuration can take effect for all instances

Instance-level configuration takes effect only for this instance

Instance-level configuration also takes effect on other instances

After the configuration is saved, you need to restart the service to take effect

Distributed detection mechanism

ACK

Watcher

Checkpoint

Interceptor

Channel processor

Channel selector

None of the aboveA. True

RDD has a lineage mechanism (Lineage)

RDDs are stored on disk by default

RDD is a read-only, partitionable distributed dataset

RDD is Spark's abstraction of underlying data

Standalone

Runtime

DataStream

FlinkCore

dfs-clear

dfs -del

dfs -rm

dfs -Is

NameNode

Data Node

ZooKeeper

JobTaoker

DataNode

RegionServer

ResourceManager

NodeManager

easy to program

good scalability

real-time computing

High fault tolerance

Used as the basis for activity streams and operational data processing pipelines

Developed by Apache Hadoop and open sourced in 2011

It has the characteristics of information persistence, high throughput, real-time, etC.

Implemented using Scala, Java language

Hive metadata storage is independent of data storage, thereby decoupling metadata and data. High flexibility, while traditional data warehouse data application is single, low flexibility

Hive is based on HDFS storage. Theoretically, the storage capacity can be expanded infinitely. The storage capacity of traditional data warehouses will have an upper limit

Since Hive data is stored in HDFS, it can ensure high fault tolerance and high reliability of data

Since Hive is based on a big data platform, query efficiency is faster than traditional data warehouses

Region

ColumnFamily

olumn

Row

HDFS

Hadoop

Memory

MapReduce

The best use case for Hive is batch jobs with large datasets

Hive can realize low-latency and fast query on large-scale data sets

Hive is built on top of Hadoop based on static batch processing, Hadoop usually has high latency andC. A lot of overhead is required for committing and scheduling

The Hive query operation process strictly follows the function execution model of Hadoop MapReduce. Hive converts the user's HiveQL statement into Map through the interpreter Reduce on Hadoop cluster

Zookeeper

Yarin

Hive

HDFS

Does not check the validity of the data

Periodically send the block-related information of this node to the NameNode

Blocks stored in different DataNodes must be different

Blocks on a DataNode. can be the same

Job Scheduler

HA Manager

Job Manager

Resource Manager

soure

Transformation

Sink

Channel

Load the file of the local path directly into the Hive table

Load the files on HDFS into the Hive table

Hive supports insert into! Single record method, so you can insert a single record directly on the command line

Insert result sets from other tables into Hive tables

1.2, 2

1, 3, 2

2, 3, 1

2, 3, 3

ZKFC (ZKFailoverController) is used as a Zookeeper cluster client to monitor the status information of NameNodel

The ZKFC process needs to be deployed in NameNodel's node and Zookeeper's Leader node

Standby NameNodej perceives the status of Active NameNodel through Zookeeper. Once the Active NameNode is down, the Standby NameNode will perform the main upgrade operation.

The ZKFC of HDFS NameNodel connects to Zookeeper, and saves the host name and other information in Zookeeper

Supervisor is responsible for resource allocation and task scheduling

Supervisort is responsible for accepting tasks assigned by Nimbus, starting and stopping Worker processes that belong to its own management

Supervisor is a process that runs specific processing logic

Supervisor is a component that receives data in Topology and then performs processing

Any machine -> same rack -> local resources

Any machine -> local resources -> same rack

Local resources -> same rack-?Any machine

Same rack->any machine->local resources