Cloudera Impala performance test - Empty cache - hadoop

I try to execute a performance test on a cloudera hadoop cluster. However, as far as Impala uses cache to store previous queries, how can I empty cache ?
Does Impala use caching?
Impala does not cache data but it does cache some table and file metadata. Although queries might run faster on subsequent iterations because the data set was cached in the OS buffer cache, Impala does not explicitly control this.
Quoted from : http://www.cloudera.com/content/cloudera/en/documentation/cloudera-impala/latest/topics/impala_faq.html#faq_performance_unique_1__faq_caching_unique_1

The file metadata caching is different from "query caching". It is just caching the locations of files and blocks in HDFS, which is something that most databases already know but Impala may not because it gets table/file metadata from Hive. The file metadata should be available to Impala in your tests.
Impala never caches queries, but file data may be cached in one of two ways:
You've enabled HDFS caching. I assume you're not doing this.
Some data read by HDFS may be in the OS buffer cache. Impala has no control over this. Some googling turns up guidance about clearing the Linux buffer caches, e.g. this unix.stackexchange.com answer.

Related

Can you use HDFS as your principal storage?

Is its reliable to save your data in Hadoop and consume it using Spark/Hive etc?
What are the advantages of using HDFS as your main storage?
HDFS is only as reliable as the Namenode(s) that maintain the file metadata. You'd better setup Namenode HA and take frequent snapshots of them, and externally store those away from HDFS.
If all Namenodes are unavailable, or their metadata storage is corrupted, you'll be unable to read the HDFS datanode data, despite those files being fine themselves, and highly available
Here are some considerations for storing your data in Hive vs HDFS (and/or HBase).
Hive:
HDFS is a filesystem that supports fail-over and HA. HDFS will replicate the data in several datanodes based on the replication factor you have chosen. Hive is build on top of Hadoop therefore can store data in HDFS as well leveraging the pros of HDFS for HA.
Hive utilizes predicates-pushdown providing huge performance benefits. Hive can also be combined with modern file formats such as parquet and ORC improving performance even more (utilizing predicates-pushdown).
Hive provides very easy access to data via HQL (Hive Query Language) which is SQL like language.
Hive works very well with Spark and you can combine them both aka retrieving Hive data into dataframes and saving dataframes into Hive.
HDFS/HBase:
Hive is a warehouse system used for data analysis therefore Hive CRUD operations are relatively slower than direct access to HDFS files (or HBase which is build for fast CRUD operations). For instance in a streaming application saving data in HDFS or HBase will be much faster than in Hive. If you need fast storage (or insert queries) and you don't do any analysis on large datasets then you should prefer HDFS/HBase over Hive.
If performance is very crucial for your application and therefore you prefer to skip the extra layer of Hive accessing HDFS files directly.
The team decides not to use SQL.
Related post:
When to use Hadoop, HBase, Hive and Pig?

Offloading unstructured data saved in RDBMS to Hadoop

My organization is thinking about offloading the unstructured data like Text , images etc saved as part of Tables in Oracle Database , into Hadoop. The size of the DB is around 10 TB and growing. The size of the CLOB/BLOB columns is around 3 TB.Right now these columns are queried for certain kind of reports through a web application. They are also written into but not very frequently.
What kind of approach we can take to achieve proper offloading of data and ensuring that the offloaded data is available for read through existing web application.
You can get part of the answer in oracle blog (link).
If data needs to be pulled in HDFS environment via sqoop, then you must first read the following from sqoop documentation.
Sqoop handles large objects (BLOB and CLOB columns) in particular ways. If this data is truly large, then these columns should not be fully materialized in memory for manipulation, as most columns are. Instead, their data is handled in a streaming fashion. Large objects can be stored inline with the rest of the data, in which case they are fully materialized in memory on every access, or they can be stored in a secondary storage file linked to the primary data storage. By default, large objects less than 16 MB in size are stored inline with the rest of the data. At a larger size, they are stored in files in the _lobs subdirectory of the import target directory. These files are stored in a separate format optimized for large record storage, which can accomodate records of up to 2^63 bytes each. The size at which lobs spill into separate files is controlled by the --inline-lob-limit argument, which takes a parameter specifying the largest lob size to keep inline, in bytes. If you set the inline LOB limit to 0, all large objects will be placed in external storage.
Reading via web application is possible if you are using MPP query engine like Impala and it works pretty well and it is production ready technology. We heavily use complex Impala queries to render content for SpringBoot application. Since Impala runs everything in memory, there is a chance of slowness or failure if it is multi-tenant Cloudera cluster. For smaller user groups (1000-2000 user base) it works perfectly fine.
Do let me know if you need more input.
Recommendation will be
Use Cloudera distribution (read here)
Give enough memory for Impala Deamons
Make sure you YARN is configured correctly for schedule (fair share or priority share) based ETL load vs Web Application Load
If required keep the Impala Daemons away from YARN
Define memory quota for Impala Memory so it allows concurrent queries
Flatten your queries so Impala runs faster without joins and shuffles.
If you are reading just a few columns, store in Parquet, it works very fast.

Impala vs Hive. How Impala circumvents MapReduce?

How is Impala able to achieve lower latency than Hive in query processing?
I was going through http://impala.apache.org/overview.html, where it is stated:
To avoid latency, Impala circumvents MapReduce to directly access the
data through a specialized distributed query engine that is very
similar to those found in commercial parallel RDBMSs. The result is
order-of-magnitude faster performance than Hive, depending on the type
of query and configuration.
How Impala fetches the data without MapReduce (as in Hive)?
Can we say that Impala is closer to HBase and should be compared with HBase instead of comparing with Hive?
Edit:
Or can we say that as classically, Hive is on top of MapReduce and does require less memory to work on while Impala does everything in memory and hence it requires more memory to work by having the data already being cached in memory and acted upon on request?
Just read Impala Architecture and Components
Impala is a massively parallel processing (MPP) database engine. It consists of different daemon processes that run on specific hosts.... Impala is different from Hive and Pig because it uses its own daemons that are spread across the cluster for queries.
It circumvents MapReduce containers by having a long running daemon on every node that is able to accept query requests. There is no singular point of failure that handles requests like HiveServer2; all impala engines are able to immediately respond to query requests rather than queueing up MapReduce YARN containers.
Impala however does rely on the Hive Metastore service because it is just a useful service for mapping out metadata stored in the RDBMS to the Hadoop filesystem. Pig, Spark, PrestoDB, and other query engines also share the Hive Metastore without communicating though HiveServer.
Data is not "already cached" in Impala. Similar to Spark, you must read the data into a large portion of memory in order for operations to be quick. Unlike Spark, the daemons and statestore services remain active for handling subsequent queries.
Impala can query HBase, but it is not similar in architecture and in my experience, a well designed HBase table is faster to query than Impala. Impala is probably closer to Kudu.
Also worth mentioning that it's not really recommended to use MapReduce Hive anymore. Tez is far better, and Hortonworks states Hive LLAP is better than Impala, although as you quoted, it largely "depends on the type of query and configuration."
Impala use "Impala Daemon" service to read data directly from the dataNode (it must be installed with the same hosts of dataNode) .he cache only the location of files and some statistics in memory not the data itself.
that why impala can't read new files created within the table . you must invalidate or refresh (depend on your case) to tell impala to cache the new files and be able to read them directly
since impala is in memory , you need to have enough memory for the data read by the query , if you query will use more data than your memory (complexe query with aggregation on huge tables),use hive with spark engine not the default map reduce
set hive.execution.engine=spark; just before the query
you can use the same query in hive with spark engine
impala is cloudera product , you won't find it for hortonworks and MapR (or others) .
Tez is not included with cloudera for exemple.
it all depends on the platform you are using

HBase - What's the difference between WAL and MemStore?

I am trying to understand the HBase architecture. I can see two different terms are used for same purpose.
Write Ahead Logs and Memstore, both are used to store new data that hasn't yet been persisted to permanent storage.
What's the difference between WAL and MemStore?
Update:
WAL - is used to recover not-yet-persisted data in case a server crashes.
MemStore - stores updates in memory as Sorted Keyvalue.
It seems lot of duplication of data before writing the data to Disk.
WAL is for recovery NOT for data duplication.(further see my answer here)
Pls go through below to understand more...
A Hbase Store hosts a MemStore and 0 or more StoreFiles (HFiles). A Store corresponds to a column family for a table for a given region.
The Write Ahead Log (WAL) records all changes to data in HBase, to file-based storage. if a RegionServer crashes or becomes unavailable before the MemStore is flushed, the WAL ensures that the changes to the data can be replayed.
With a single WAL per RegionServer, the RegionServer must write to the WAL serially, because HDFS files must be sequential. This causes the WAL to be a performance bottleneck.
WAL can be disabled to improve performance bottleneck.
This is done by calling the Hbase client field
Mutation.writeToWAL(false)
General Note : Its general practice that while doing bulkloading data, WAL is disabled to get speed. But side effect is if you disable WAL you cant get back data to replay if in case any memory crashes.
More over if you use solr+ HBASE + LILY, i.e LILY Morphiline NRT indexes with hbase then it will work on WAL if you disable WAL for performance reasons, then Solr NRT indexing wont work. since Lily works on WAL.
please have a look at Hbase architecture section

Extracting Data from Oracle to Hadoop. Is Sqoop a good idea

I'm looking to extract some data from an Oracle database and transferring it to a remote HDFS file system. There appears to be a couple of possible ways of achieving this:
Use Sqoop. This tool will extract the data, copy it across the network and store it directly into HDFS
Use SQL to read the data and store in on the local file system. When this has been completed copy (ftp?) the data to the Hadoop system.
My question will the first method (which is less work for me) cause Oracle to lock tables for longer than required?
My worry is that that Sqoop might take out a lock on the database when it starts to query the data and this lock isn't going to be released until all of the data has been copied across to HDFS. Since I'll be extracting large amounts of data and copying it to a remote location (so there will be significant network latency) the lock will remain longer than would otherwise be required.
Sqoop issues usual select queries on the Oracle batabase, so it does
the same locks as the select query would. No extra additional locking
is performed by Sqoop.
Data will be transferred in several concurrent tasks(mappers). Any
expensive function call will put a significant performance burden on
your database server. Advanced functions could lock certain tables,
preventing Sqoop from transferring data in parallel. This will
adversely affect transfer performance.
For efficient advanced filtering, run the filtering query on your
database prior to import, save its output to a temporary table and
run Sqoop to import the temporary table into Hadoop without the —where parameter.
Sqoop import has nothing to do with copy of data accross the network.
Sqoop stores at one location and based on the Replication Factor of
the cluster HDFS replicates the data

Resources