Huawei HCIP-Big Data Developer (H13-723) Certification Sample Questions

Huawei H13-723 VCE, HCIP-Big Data Developer Dumps, H13-723 PDF, H13-723 Dumps, HCIP-Big Data Developer VCE, Huawei HCIP-Big Data Developer PDF Getting knowledge of the Huawei H13-723 exam structure and question format is vital in preparing for the Huawei Certified ICT Professional - Big Data Developer certification exam. Our Huawei HCIP-Big Data Developer sample questions offer you information regarding the question types and level of difficulty you will face in the real exam. The benefit of using these Huawei H13-723 sample questions is that you will get to check your preparation level or enhance your knowledge by learning the unknown questions. You will also get a clear idea of the exam environment and exam pattern you will face in the actual exam with the Huawei Certified ICT Professional - Big Data Developer Sample Practice Test. Therefore, solve the Huawei HCIP-Big Data Developer sample questions to stay one step forward in grabbing the Huawei Certified ICT Professional - Big Data Developer credential.

These Huawei H13-723 sample questions are simple and basic questions similar to the actual Huawei HCIP-Big Data Developer questions. If you want to evaluate your preparation level, we suggest taking our Huawei Certified ICT Professional - Big Data Developer Premium Practice Test. You might face difficulties while solving the real-exam-like questions. But, you can work hard and build your confidence on the syllabus topics through unlimited practice attempts.

Huawei H13-723 Sample Questions:

01. A reporting web service on an MRS security cluster authenticates to Kerberos with one service keytab. It must read HDFS files on behalf of hundreds of signed-in analysts so that HDFS permission checks and audit logs record each analyst, not the service. The security team will not distribute analysts' keytabs or passwords to the service.

Which measures together meet these requirements? (Choose two.)

a) Create a proxy UserGroupInformation for each analyst from the service's login user and run HDFS calls in its doAs block

b) Keep each analyst's keytab on the service host and log in as that analyst for each request

c) Read every file as the service user and enforce each analyst's access rules in the service's own code

d) Authorize the service user as a proxy user on the cluster, restricted to the hosts and groups it serves

e) Run HDFS calls in a doAs block of the service's own login user, passing the analyst name as a parameter

02. A Java application on a FusionInsight HD security cluster already writes to HDFS after calling UserGroupInformation.loginUserFromKeytab. The developer adds a KafkaConsumer to the same process, with the SASL Kerberos security protocol configured, but the consumer fails at startup with a login error stating that no JAAS configuration was found.

What must the developer add so that the Kafka client can authenticate?

a) The Kafka broker's service keytab, copied into the directory the application runs from

b) A second call to loginUserFromKeytab, placed just before the consumer is created

c) A JAAS KafkaClient entry naming the user's keytab and principal

d) The cluster's hdfs-site.xml and core-site.xml files on the application classpath

03. A telecom operator stores call detail records in HBase keyed by a call ID that increases with every call, and writes concentrate on one region. The team compares three remedies: a prefix of hash(callID) mod N, a random salt prefix between 0 and N minus 1, and reversing the digits of the call ID. Billing needs single-call lookups by call ID and occasional range scans over consecutive call IDs.

Which statements about these remedies are correct? (Choose two.)

a) With a random salt prefix, a lookup for one call ID goes straight to one region without trying other prefixes

b) With the hash-mod-N prefix, a lookup computes the prefix from the call ID, but a range of consecutive IDs needs N scans

c) Reversing the digits keeps consecutive call IDs next to each other, so a range of IDs remains one contiguous scan

d) Replacing the call ID with its full MD5 hash keeps a range of consecutive IDs readable in one scan

e) Reversing the digits spreads sequential writes, and a lookup by exact call ID is still a single Get

04. Two hundred instances of a billing service compete to be the single active instance. Today each instance tries to create a fixed ephemeral znode named /billing/leader and watches it when creation fails. Whenever the leader crashes, all instances are notified at once and flood the ZooKeeper ensemble with create requests.

Which design keeps automatic failover while removing this burst of requests?

a) Use ephemeral sequential znodes where the lowest one leads and each instance watches only its predecessor

b) Make /billing/leader a persistent znode so that it survives the end of the leader's session

c) Have each instance watch the children of /billing instead and retry the create after a random back-off delay

d) Have each instance create a persistent sequential znode and treat the lowest number as the leader

05. A multi-tenant SaaS product stores every tenant's documents in one Elasticsearch index and indexes each document with custom routing on tenant_id. Most tenants are small, but one tenant holds about a third of all documents and generates most of the search traffic.

Which statements about this design are correct? (Choose two.)

a) Routing lets the primary shard count grow later without reindexing, because tenants can be remapped

b) A GET by document ID still finds the document reliably when the routing value is left out

c) A search that passes the tenant's routing value is sent to one shard instead of to every shard

d) The large tenant's documents all land on one shard, which becomes oversized and hot

e) Routing spreads each tenant's documents evenly across all primary shards of the index

06. A Spark SQL job joins a large fact table of shipments with a carrier dimension table. When the job was written the dimension table was small, and a developer added a broadcast hint. The dimension table has since grown to tens of gigabytes after carrier history was added, and the job now fails with out-of-memory errors on the driver and executors during the join.

What is the most appropriate fix?

a) Cache the dimension table with persist() before the join so the broadcast reads it from executor memory instead of HDFS

b) Remove the hint and let Spark use a shuffle-based sort-merge join

c) Collect the dimension table to the driver first and build a lookup map that each task reads through a closure

d) Increase the number of shuffle partitions so the broadcast table is divided into smaller pieces per executor

07. A product-detail API uses cache-aside: look up Redis, and on a miss read the database and fill the cache. A bot now repeatedly requests a few thousand product IDs that do not exist. Each such request misses Redis, because nothing is cached for a missing product, and reaches the database, whose load has tripled. The hit rate for real products is unchanged.

Which measures stop these requests from reaching the database? (Choose two.)

a) Cache a placeholder for a missing ID with a short TTL, so that repeat lookups of it are answered by Redis

b) Add random jitter to every cache TTL, so that cached product entries no longer expire at the same moment

c) Guard each cache rebuild with a mutex, so that only one request per missing key queries the database

d) Raise maxmemory, so that more product entries stay cached and fewer requests fall through to the database

e) Check each ID against a Bloom filter of existing product IDs, and reject the IDs that it rules out

08. An online profile service performs random Get calls on an HBase table and meets its latency target most of the day. Every night a reporting job runs a full-table Scan over the same table, and for hours afterward the service's latency rises and the BlockCache hit ratio drops sharply before slowly recovering.

Which change protects the profile service?

a) Enlarge the MemStore so that profile Gets are answered from memory

b) Run a major compaction after the report, so the HFiles are reorganized

c) Raise the reporting Scan's caching so the job finishes in fewer RPCs

d) Have the reporting Scan call setCacheBlocks(false), so blocks it reads only once do not evict the service's hot blocks

09. A Redis instance holds shopping-cart state for a busy retail site. The business accepts losing at most about one second of cart updates if the server crashes, the write rate is high enough that a disk sync on every command would cut throughput too far, and restarts should still load quickly.

Which persistence configuration fits best?

a) RDB snapshots only, taken frequently, because each snapshot captures every write since the previous one

b) AOF with appendfsync everysec, rewritten with an RDB preamble so that restarts load fast

c) AOF with appendfsync always, because syncing each write is the only way to bound loss to one second

d) No persistence, relying on a replica, because replication makes every acknowledged write durable

10. A Flink job computes hourly event-time revenue per store. Finance needs each hour's figure as soon as the watermark passes the hour, corrected if stragglers arrive within the next 15 minutes. Records later than that must not be lost; they go to a separate audit table. The watermark strategy already fits the normal disorder.

Which window configuration meets all three requirements?

a) Keep the watermark, set allowedLateness to 15 minutes, and route later records with sideOutputLateData

b) Raise the watermark's out-of-orderness bound by 15 minutes, so stragglers are already inside each window

c) Set allowedLateness to 15 minutes; Flink then writes later records to a late table

d) Use processing-time windows, so that no record is late, and produce the audit copy from a second job that rereads the topic each hour

Answers:

Question: 01 Answer: a, d	Question: 02 Answer: c	Question: 03 Answer: b, e	Question: 04 Answer: a	Question: 05 Answer: c, d
Question: 06 Answer: b	Question: 07 Answer: a, e	Question: 08 Answer: d	Question: 09 Answer: b	Question: 10 Answer: a

Note: For any error in Huawei Certified ICT Professional - Big Data Developer (H13-723) certification exam sample questions, please update us by writing an email on feedback@certfun.com.