Also, you may consider checking a graph that we posted comparing the
performance of BookKeeper with the one of HDFS using a local file
system and local+NFS in the jira issue 5189 (https://issues.apache.org/jira/browse/HADOOP-5189
).
-Flavio
On Feb 20, 2009, at 10:05 AM, Flavio Junqueira wrote:
Hi Bill, I'm sorry, I missed this message initially. I'm sending
below a table that gives you throughput figures for BookKeeper. The
rows correspond to distinct BookKeeper configuration (ensemble size,
quorum size, entry type), and the columns to different values for
the length of an entry in bytes. The throughput values correspond to
one client writing 400K records (we call them entries)
asynchronously to a ledger. Finally, the table shows write
throughput in thousands of operations per second.
128 1024 8192
3-2-V 32.80 26.45 5.89
4-2-V 41.72 31.53 6.55
5-2-V 46.89 32.45 6.61
4-3-G 28.02 21.61 4.37
5-3-G 34.91 28.22 4.60
6-3-G 41.22 31.70 4.55
Let me know if you have more questions, I appreciate your interest.
Thanks,
-Flavio
On Feb 14, 2009, at 2:56 PM, Bill de hOra wrote:
Patrick Hunt wrote:
A bit about BookKeeper: a system to reliably log streams of
records. In BookKeeper, servers are "bookies", log streams are
"ledgers", and each unit of a log (aka record) is a "ledger
entry". BookKeeper is designed to be reliable; bookies, the
servers that store ledgers can be byzantine, which means that some
subset of the bookies can fail, corrupt data, discard data, but as
long as there are enough correctly behaving servers the service as
a whole behaves correctly; the meta data for BookKeeper is stored
in ZooKeeper.
Hi Patrick,
this sounds cool. Are there any figures on throughput, ie how many
records BookKeeper can process per second?
Bill