I'm so excited that they finally open-sourced this. It's relatively old tech at Yahoo, stuff folks outside never got to see. It was difficult to explain to later colleagues the stuff I knew about shared-memory databases because I couldn't give them a frame of reference.<p>mdbm performance is even better on FreeBSD than Linux because FreeBSD supports MAP_NOSYNC, which causes the kernel not to flush dirty pages to disk until the region is unmapped. Perhaps mdbm's release will finally get the Linux kernel team to provide support for that flag.
I'm not very comfortable with storage engines that directly build on memory mapped files. MongoDB's current storage engine is mmap based and it's sub optimal at best which is undoubtedly part of the reason they're building a completely new storage engine now (WiredTiger).
I'm curious to see what the total timings would be like to get the data in a useable form - as opposed to just fetching a record from a data store. As noted - these data stores just store and retrieve data and don't do things like joins or ordering, etc.<p>Could there be a comparison between these datastores and the traditional ACID compliant databases when it comes to retrieving actual data in a useful format? E.g. perhaps doing a join or an ordering of some sort? I don't expect databases (e.g. Oracle, MS SQL Server, DB2) to be faster in raw performance, but I do expect them to be faster in terms of total development time and bug fixing since the application developer wouldn't have to do the locking, page pinning/unpinning, etc. manually.
Could not install this in Ubuntu 12.04 - basic commands are failing. I think they tested only in BSD?<p>ln -s -f -r /tmp/install/lib64/libmdbm.so.4 /tmp/install/lib64/libmdbm.so
ln: invalid option -- 'r'
Try `ln --help' for more informatio
Thoughts on using this as a cache instead of memcache or redis? Yes, it does not have nearly as many features or functions but when raw performance is needed I could see this working (given an api for using this via Node.JS, PHP, etc.).
Do people get annoyed by all the JavaScript frameworks and Databases coming out in regards to adoption from a company point of view? I mean every other day a new database comes out and claims to be better in one way or another than something else and then its like "fuck I picked X when now there's Y"<p>It seems over the last year technology has been growing more rapidly than any other period.<p>Fun times but so hard to keep track of everything!