That approach will tend to hot-spot on ingest: the latest updates will always go to the first tablet. Maybe that's ok, but you might want some scheme to spread those updates to other tablets.
-Eric On Tue, Apr 23, 2013 at 8:12 PM, Venkat <[email protected]> wrote: > Thanks Keith, This is a great suggestion. Trying out the reverse of that > .. (999999999999 - timestamp)+machineID > > Thanks ! > > > On Tue, Apr 23, 2013 at 7:59 AM, Keith Turner <[email protected]> wrote: > >> >> >> >> On Mon, Apr 22, 2013 at 8:59 PM, Venkat <[email protected]> wrote: >> >>> It looks like versioningiterator gives me the past versions for a cell >>> .... but I need to look across rows .. >>> >>> I am inserting time series data for set of machines into accumulo and i >>> am using a combo of machineid + timestamp as the row ID and metric as the >>> cq. So for me to get the last time a metric has been inserted for an >>> asset, i need to look across row IDs and get the last 5 rows that got >>> inserted with that metric ... >>> >> >> You could consider sorting your data such that the most recent timestamps >> come first. If you insert (machineid + (999999999999 - timestamp)). To >> get the 5 most recent row, start scanning at (machineid + 999999999999) and >> stop when you have 5 rows. I picked a completely arbitrary number of 9's >> for the example, not sure what time resolution its appropriate for. >> >> >>> >>> >>> >>> On Mon, Apr 22, 2013 at 2:48 PM, Billie Rinaldi < >>> [email protected]> wrote: >>> >>>> You could use a VersioningIterator with maxVersions set to 5. >>>> >>>> Billie >>>> >>>> >>>> >>>> On Mon, Apr 22, 2013 at 11:29 AM, Venkat <[email protected]> wrote: >>>> >>>>> hello accumulo, >>>>> >>>>> what is a good way to get rows in chronological order ? i am trying >>>>> to get to a use case of "give me the last 5 inserted for this particular >>>>> cf, cq ..." timestampfilter works well when I know the times but i dont >>>>> know the exact times of insertion. >>>>> >>>> >>>> >>> >> >
