Weekly Edition
Return to the Development page
| |||||||||||||
Hibari "big data" database released
Gemini Mobile Technologies has sent out a
press release announcing the availability (under the Apache license) of
"Hibari," a non-relational database, implemented in Erlang. "Hibari
is a database optimized for the highly reliable, highly available storage
of massive data, so-called 'Big Data.' Hibari can be used in Cloud
Computing Applications such as web mail, Social Networking Services (SNS),
and other services requiring storage of tera-bytes and peta-bytes of new
daily data."
(Log in to post comments)
"high-level" is always a euphemism Posted Jul 27, 2010 21:06 UTC (Tue) by ncm (subscriber, #165) [Link]
Instead of writing "implemented in Erlang", why not just write "slow"?
Seriously, how else could the implementation language matter? Aside from speed, either the thing is useful or it ain't. I appreciate the warning. However, reading the press release, it seems as if the event being announced is not just releasing a new version, but actually freeing up the code. People who hope to customize it probably do need to know what language they'll be coding.
Speaking of Erlang, here's the funniest thing you'll see today:
"high-level" is always a euphemism Posted Jul 27, 2010 22:37 UTC (Tue) by NAR (subscriber, #1313) [Link] Seriously, how else could the implementation language matter?For example with Erlang you get the possibility of hot code loading, which might mean upgrade without downtime. Also with Erlang distribution is built into the language/VM, so it might be very simple to scale the application by adding extra hardware. These features could be useful for a database of "big data". Of course these are only possibilities, it depends on the application too. Also, any language that is not C/C++ implicates that you won't get segmentation faults because of an off-by-one error or a buffer overflow (although I have to add I've seen both the Java and the Erlang VMs segfaulting).
"high-level" is always a euphemism Posted Jul 27, 2010 23:24 UTC (Tue) by njs (guest, #40338) [Link]
This application is not going to be CPU bound, so who cares how slow Erlang is? It'll be IO bound (for which language doesn't matter), and scalability bound (for which Erlang, all else being equal, is going to kick the pants off pretty much any other mature/widely deployed language. It's a sad commentary on the current state of programming languages that every otherwise attractive general purpose language sucks mightily at distributed computing, but there it is).
Hibari "big data" database released Posted Jul 28, 2010 12:13 UTC (Wed) by SEJeff (subscriber, #51588) [Link]
My first question would be, "How does this compare to Riak?" which is a nosql database which is written in erlang and has been around for awhile now.
Hibari "big data" database released Posted Jul 29, 2010 17:13 UTC (Thu) by dmag (subscriber, #17775) [Link]
> How does this compare to Riak?
Riak is AP of CAP, where as Hibari appears to be CA. Both are KV-stores, which means you can't query by values, only keys. (Although Riak has map-reduce, wherein you process *all* key/values.)
(One thing to be careful of: Riak has multiple storage backends, so benchmark results can differ wildly depending on which you choose.)
> a nosql database which is written in erlang and has been around for awhile now
That also describes CouchDB, for what it's worth. (Not much. Organizing database projects by language isn't useful. Writing something in Erlang doesn't make it scalable, distributed or multi-core any more that writing it in C.)
|
|||||||||||||