Amazon CloudSearch vs. HP Autonomy Intelligent Universal Search (discontinued)

Amazon Cloudsearch can be suitable for some queries that require fast data. For example, in our case, we used CloudSearch, in a tool called Global Search. That will search everything like names, emails and a lot of stuff in our application. If you want fast data and you have a simple query, Global Search isn't appropriate for you.

Incentivized

It does a decent job at its core functions (that other free software does just as well or better).

• Really fast queries
• Good Reporting
• Reduce the cost of the server

Incentivized

• It adheres to traditional Microsoft standards such as: fact-dump documentation with no coherent story or 'best practices' information, inability to automate common tasks, intentional obfuscation of its basic operations.
• It provides OK search results. Not great, but OK.

• Can take some time to implement
• The initial configuration can be tricky
• Takes some time to update the values

Incentivized

• There are about a dozen different config files to maintain, and the most important one is dynamically modified by Autonomy itself while it runs. Which means that it is impossible to automate the configuration or keep the configs in versioned source control. Even `cp *.cfg ~/cfgbak/` won't help you roll back a change, because it is never safe to restore a previous config. You'll be using `diff new.cfg old.cfg` a lot.
• The Linux port is poorly thought out. The binaries are named *.exe. The StartService.sh scripts contain both `echo 'Are you sure you want to start the service? Hit ctrl-C to cancel''; read dummy` and, I kid you not, a `chmod a+x /path/to/my/binary.exe`.
• Many features are poorly documented, leading to lots of back and forth with the support department just to answer basic questions like "what does this error code in my logs signify?"
• It seems to reinvent the wheel, poorly, everywhere. E.g. the scheduled backup feature rolls through a user-defined finite list of directories in which to store backups. On day 0 it uses directory 0, on day 1 it uses directory 1, and after day N it rolls back and overwrites directory 0. Why would this be preferable to using a single directory and naming zip files based on the current timestamp?

Management wants to see ROI on the (hefty) cost of purchasing this software, and has mandated that we continue using it. We would prefer to switch immediately.

I didn't investigate the best alternatives to CloudSearch, but did help with implementing this feature in our application. But from what i tested and used - Cloudsearch is very fast to get queries. Some negative points can be the time to implement this and some configurations that can be tricky.

Incentivized

• Positive Point - Reduced the server load
• Negative point - Not suitable for all queries
• Negative Point - Time to implement this feature

Incentivized

• I have learned to tack a zero onto the end of any estimate I make for how long an Autonomy change will take in both planning and implementation.