This is normal for all data systems. The speed of random reads is a test of disk seek speed (typically 5-9ms) and how much locality your random reads are, in other words, how effective your cache is.
On Dec 11, 2009 11:48 AM, "ruslan usifov" <ruslan.usi...@gmail.com> wrote: Hello! I'm new in Hbase and misunderstand some things. Configuration: 3 Ubuntu 9.10 (ip6 off) servers (srv1, srv2, srv3) Hbase 20.2 HadoopCore 20.1 Servers are connected with Gigabit Ethernet switch to monitor hardware i use: nmon (to monitor cpu, discs) iftop (to monitor network) my Hadoop configuration 3 Datanodes (on each server) 1 Namenode (srv1) 2 Zookeper (srv2, srv3) 3 RegionServer (on each server) 1 Hmaster (srv1) Replication factor is 2 I have table with 20.000.000 (20 millions) rows in it, in each row i have two columns (in each column is about 10 bytes of data). When i try to random read this table i have slow(i think) reads, about 700 requests per second (if i use one read thread), but no cpu no discs are used full (cpu about 15%, discs also doesn't have much activity) I use follow code to retrieve records: Get get = new Get((rand.nextInt(20000000) + "").getBytes()); Result res = htable.get(get); Is this read behaviour normal? Why they so slow?
