Information Technology Reference
In-Depth Information
(4)
IT Manager edits Cassandra's file content (cassandra-topology.properties), then
sets the names of the data center and the storage location of the nodes (rack
number).
(5)
IT manager edits Cassandra's file content (cassandra.yaml), and then changes the
content of endpoint_snitch [16] to PropertyFileSnitch (data center management
mode).
(6)
IT manager executes command (create keyspace test with strategy_options
={DC1:2,DC2:1} and placement_strategy='NetworkTopologyStrategy') in
Cassandra's primary data center, and then creates a form and initialize remote
backup.
(7)
IT manager eventually has to test the performance of writing, reading, and offsite
data backup against large amounts of information using Thrift Java.
4.2
Intelligent Parameter Adaptation
In order to maintain the smooth remote backup over internet, we focus on the quality
of services (QoS) about the data transmission and receipt. In other words, we have to
deal with the crucial problems about jitter, loss, latency, and throughput. This paper
introduces an intelligent adaptation for tuning data transmission and receipt
parameters appropriately in both datacenters. As shown in Fig. 1 the diagram
illustrates an intelligent adaptation using an adaptive network-based fuzzy inference
system (ANFIS) [8] along with particle swarm optimization (PSO) [9] approach,
where C, D, W, R, and M denote the normalized CPU clock rate, the size of SDRAM,
network bandwidth, the ratio of read to write operation, and the number of message,
respectively, to adjust remote backup parameters in HBase, i.e., the time interval
between RegionServer and master as well as the ratio of read to write operation in
system. It is noted that PSO approach is utilized to search the optimal weight
parameters of ANFIS. Besides, the same scheme of ANFIS is applied for tuning
remote backup parameters in Cassendra, namely the maximum size of commitlog
space in memory, the maximum throughput in read/write operation, and the
percentage of cache release for saving the out of bound of Java heap as shown in Fig.
2 where N represents the number of read plus write operation instead of R in Fig. 1.
We have collected a lot of data by the manner of trial-and-error during the
experiments. Once the data collection has completed, those of data have been put into
the ANFIS for training and validating so that we can get a trained ANFIS system for
infer the key parameters such as
(1) hbase.regionserver.msginterval,
(2) hbase.ipc.server.callqueue.read.ratio,
(3) commitlog_total_space_in_mb,
(4) compaction_throughput_mb_per_sec, and
(5) reduce_cache_capacity_to.
After that, the remote data center backup has been tested in the cloud computing
system based on web interface and as a result it performs very well on HBase and
Cassendra data center backup remotely.
Search WWH ::
Custom Search