One thing that could improved upon is the level of concurrency. The documentation for this solution could also be improved. 

Customer support can be improved because if we have any kind of problem and raise a ticket, there's a lag in response time. I've also found that there is some wastage because when we provision the clusters as per the request from the application team, they provide too much memory, CPUs and displays.  Couchbase doesn't use that much memory or CPU. Maybe they can do some kind of capacity planning to solve that. I'd like to see our application team have non-DBA access so that the DBAs work with the application team for a minimal amount of time. 

Their security model leaves something to be desired. Their indexes could use some work, but once we found out what was going on with them, it became second. So, the biggest one so far is integrating with an external security model and having a security model that you can use. I know there's a performance hit with it, but if possible, I'd like to be able to go down to potentially a column-level instance. I'd like to be able to set up entitlements based on that. I can do that currently, but it's a little bit convoluted in the way I have to do it. It can become, from a large enterprise standpoint, a bit unwieldy in the way I have to manage it to get it to work.

The other thing that they could really improve is their documentation. Their online documentation or just documentation in general is pretty bad. The online documentation is very confusing. It's not very intuitive to find information. There are no white papers on best practices for doing things. As much as everybody rags on IBM, considering the various different software products that IBM has, their documentation, for the most part, is very good. You can find many white papers, and it's very easy to find things, whereas, with Couchbase, the whole documentation setup is very confusing. It's not usable, for the most part.

They should support native S3 ingestion into the default bucket or into the system. Currently, when I want to look at S3 stuff, I use the analytics engine, and that works fine. However, when I want to use S3 as a landing point to be able to load into the default Couchbase—not use in the analytics engine but just to import into Couchbase—there's no native way of having Couchbase talk to S3 to get JSON documents out and loaded into the data store. So, I've to use a third-party product called Data Sync to be able to pull stuff out of S3. The integration options aren't as robust as they probably need to be. They only have about two.

It's easy to deploy. Where the challenge comes in is when you start putting data in, doing the indexes, and doing the integration with systems. Integration is one of their weakest points. Natively, there should be a wide range of integration options to be able to get data in. Once you get the data in, it's really easy to get the data out. It's very simple. It's a JDBC, etc. It's just that the integration to get the data in can be a challenge.

The scripting language for this solution could be improved. A big selling point is that they're like SQL server but there is still quite a lot of missing functionality. They are slowly adding additional functionality but without it, the solution ca be complex to use. 

Secondly, this solution does not function well on a Windows based server or infrastructure. Hosting it on a Windows server can max out your CPU and bring down your cluster. This is one of the main reasons that we stepped away from having that in this kind of environment as well as needing to buy a license for support. There is also a lack of documentation. 

We would like to have a better management of Kubernetes with the free, open source version of Couchbase. We don't have any major complaints other than that.

Most of the time, the only issued that I have faced was when we upgraded the Couchbase version. It is very difficult to load the backup of the older version to the newer version.

The only problem with it, I mentioned, the loading data from a different version. For that, we use replication to solve the issue. This is the only bad issue we faced, otherwise it is perfect.

The failover and failback could be a bit easier. When I looked at it last time, it had to be manually done. It also took over an hour for us to rebalance all the nodes.

Couchbase could improve the design of the UI because it should be optimized for viewing statistics or a similar feature.

There are some limitations to the database. The SQL database cannot handle real-time processing for critical IoT scenarios. What we have to do is store our data into the database then code it out, this wastes a lot of time.

I have tried multiple libraries in a demo they provide and it works fine, but when it merges with libraries, it creates a problem.