[Neo-users] Current tests

[Vincent Pelletier]

Another quick update.

Today, we reached one week uptime with NEO.

Now, this statement must be taken with some salt: In our test setup, there is 
one MySQL wihch is used by the application running on Zopes as a relational 
index for ZODB (ie, nothing to do with NEO itself). That MySQL half-crashed 
twice in that uptime, causing huge performance hit on Zope level (mysql still 
responded, just very very very slowly), which caused NEO to be under very 
little stress. Zopes & involved MySQL were restarted twice.

Once this grain of salt taken into account, most part of NEO itself (master & 
storage nodes) were kept running for the whole duration, and their own MySQL 
didn't suffer from that bug - so those process do have 1 week uptime today.
On site availability level, contiguous uptime is of 4 days, and there were 2 
chunks of around 12 hours before that (MySQL crashed twice very early in last 
week).

Some raw numbers:
- ~50M rows in "obj" table, for 21M different objects (difference being the
  number of their non-current revisions)
- 2.6M transactions
- 650GB of logs
  Traceability for the win !
  Of course they are useless in actual production environment (...when NEO
  will be production-ready).

Document creation speed is 5400 per hour, with a decrease rate of ~3 
documents per hour less per hour, which means each document created slows the 
next creation by about 0.05%. For now, this decrease looks linear. We hope 
that it is either logarithmic (and will reduce with time) or caused by 
external factors (eg. logs piling up in disk). More uptime and measures will 
tell.
From older measures, we saw that when logs are disabled, we reach more than 
6100 documents/hour.

-- 
Vincent Pelletier