Since Cloudera acquired HDP, it's been bundled with CBH and HDP. However, the biggest challenge is cloud storage integration with Azure, GCP, and AWS. These platforms offer competitive storage solutions like Gen2, Gen1, Bigtable, BigQuery, Lightstore, S3 buckets, etc., which pose a significant competition to HDP.

The cost of the solution is high and there is room for improvement.

We'd like the solution to be easier.

There used to be a free version on offer. Now, everything is paid.

It's not quite as easy to install as EMR.

It's at end of life and no longer will there be improvements. 

Hortonworks Data Platform has a limited user community. I haven't seen much discussion about user experiences. More information could be there to simplify the process of running the product.

We are happy with the platform but we are also looking at what else is out there. We are comparing what other teams are using in our company to the solution we have adopted.  

The main difference, and what seems better with other tools, is the deployment. That part is not entirely clear to us. We can create models, but once we create the models we are not sure as to how to deploy them.  

The version control of the software is also an issue.  

Maybe these improvements are already included in the newest release but we are not aware of it because nobody on our team has had the opportunity to try it. I do not know how well this product is supported.  

Integrate a complete hive web client (Ambari views), like Hue Today, in the next release.

I work a lot with banking, IT and communications customers. Hortonworks must improve or must upgrade their services for these sectors.

Each customer has different requirements. From the IT side, someone who has some experience of the cluster, computer clustering, computer networking, different fire defense, for example, it is so important that they have some additional graphics, some additional service reports added to the Hortonworks current user interface, which could provide easy images. Especially if they are using it without any experience first.

We have had problems with the backup and with services that require a disaster site. We are still struggling with some of these issues.

We are having trouble with Active Directory and Hive integration.

I would like to see more support for containers such as Docker and OpenShift.

I would like to see an increase in usability for the Apache Storm engine within the data platform.

The NiFi platform could be enhanced. This refers to the data ingestion in a workflow. 

It would also be nice if there was less coding involved. 

File system to provide indexed access to individual records with in-place update/delete. Also, Security integration through a common interface for authentication, authorization, disk encryption, network encryption, data access layer, data masking, etc.

Hadoop, does not provide improved performance, compared to traditional RDBMS, unless processing batches in the TB-PB range, or if the Hadoop platform has significantly more resources available.

• Stability
• It needs to be more mature
• Security
• User friendliness

• Rolling upgrade
• Disaster recovery features such as mirroring should be supported

• Cost
• Reliability
• Speed
• Ease of use

It needs a better UI.

The ability to update data is an area where the product could improve.

• Security
• Performance
• Workload management

There have been some governance initiatives, but they are far from production ready. I would like to see a big improvement in that space, as governance is critical in many regulated industries.

It would be interesting to have an easy way to implement multi-tenant for HDFS with federation. At the moment, you have to do it manually in command line.

Also, it needs to support having more than two HDFS namenodes. HDFS supports more than 2 namenodes, but Hortonworks doesn't.

As this is open source, there are certain features that need tuning, such as the Hue UI. More stability on this would be helpful.

Security- Although they support Knox and Ranger and Kerberos, they are still missing attribute-level encryption features.

Name Node High Availability is still not stable (memory issues).

It needs to be quicker and to have the ability to automate deployment on multiple nodes.

Hive performance. If Hive performance increased, Hadoop would replace (not everywhere) traditional databases (Oracle/Teradata, etc.), which would save a lot of money for the company.

Deleting any service requires a lot of clean up, unlike Cloudera.