MemSQL After installation

After basic installation of MemSQL, Now we need to understand the important parameters that effects the cluster. The three main parameters that affects the cluster are

• durability
• redundancy_level
• master-aggregator

The above three parameters should be added in master aggregator at /var/lib/memsql/memsql.cnf and after adding the above parameters restart the server using "service memsql restart" command

USE & IMPACTS

###############

• durability : - This is very important feature of MemSQL. If this is on then MemSQL takes periodic snapshot and log changes to disk and if its off then it will save only metadata to disk. This is the beauty of MemSQL that your data is secure on disk as well as available in-memory for high-availability. 

• redundancy_level : - To add redundancy to the database we can set redundancy_level which means data is available in separate groups called availability groups which contains a copy of every partition in the system- some as masters and some as slaves. Currently MemSQL supports two level of redundancy level 1 and 2 which means each partition of database on master has a copy of slave on other leaf.

• master-aggregator : - MemSQL is actually two tiered clustered architecture which has two types of server. Aggregator and the Leaves. Aggregators are the nodes which acts as mediator between client request and the cluster nodes (leaves). Aggregators only contains metadata and the leaves contains actual data. Aggregators are of two types Master and the Child aggregaors. To make master aggregator one should have change in memsql.cnf file is "master-aggregator" and restart the MemSQL.

TO BE CONTINUED............................

MemSQL Installation

MemSQL installation is very simple. But before installation its important to choose the suitable package. Earlier MemSQL subscription was free for one month but as per my knowledge they have extended the trial period upto 3 months. So its very convenient for end user to evaluate the product with time. No need to hurry.

As per my experience MemSQL has separate subscription for trial and for production. The production subscription has lot more as compare to trial subscription. So its better to go for easiest installation option for trial.

One of the common industry installation options is rpm. In this entire blog I will discuss the things based on rpm installation.

How do we start rpm installation (step wise)

1. REGISTER ON MEMSQL SITE www.memsql.com
2. YOU WILL GET MAIL FROM MEMSQL LIKE BELOW
3. Name: Jeetendra Ranjan                  Trial License Key: 91b7328a50e748c89bfa8702d55c1111
                     License File: Click here to download
                             now go to the bottom of the page and you will see the rpm installation link like below
   
   The following installation options require installing MemSQL on each node of your cluster. For instructions on installing MemSQL, click here.
   MemSQL Download Links
   $ wget http://download.memsql.com/91b7328a50e748c89bfa8702d55c1111/memsql-3.2.x86_64.rpm
   Here I would like to share one more link which you will see. this is memsql-ops. So memsql-ops is very useful monitoring tool for memsql as this is very difficult to monitor all the servers(aggregators and leaves individually. So go for memsql-ops installation also. here is the link
   $ wget http://download.memsql.com/ops-3.1.1/memsql-ops-3.1.1.x86_64.rpm
4. Before we start installation few very important point. First is MemSQL does not support 32-bit architecture so don't try to install on 32-bit servers. Secondly as we know MemSQL is two tiered architecture so same installation should be done on all the aggregators as well as on leaves. Very important is MemSQL does not have their own client tool so it support MySQL as a client. So on all the servers MySQL client should be installed.
5.  [root@localhost opt]#wget http://download.memsql.com/91b7328a50e748c89bfa8702d55c1111/memsql-3.2.x86_64.rpm
6.  [root@localhost opt]# ls
     memsql-3.2.x86_64.rpm
7. 4. [root@localhost opt]# rpm -i memsql-3.2.x86_64.rpm
    [root@localhost opt]# rpm -qa  grep -i memsql
    memsql-3.2-1.x86_64
    [root@localhost opt]# rpm -qa  grep -i mysql (i
    MySQL-client-5.5.27-1.linux2.6.x86_64
8. add below at memsql.cnf (/var/lib/memsql) at aggregator
   
    port  = 22407
    redundancy-level=2
    master-aggregator
9. [root@localhost opt]# service memsql start
10. mysql -u root --password='' -h 127.0.0.1 -P 22407 --prompt="memsql> "
11. memsql> add leaf 'root'@'192.168.2.2':23976; (at aggregator)
     memsql> add leaf 'root'@'10.20.1.3':23976; (at aggregator)
12. memsql> show leaves;
    +---------------+-------+--------------------+---------------+-----------+--------+--------------------+---------------------------+
     Host           Port   Availability_Group  Pair_Host      Pair_Port  State   Opened_Connections  Average_Roundtrip_Latency
    +---------------+-------+--------------------+---------------+-----------+--------+--------------------+---------------------------+
     192.168.2.2  23976                   1  10.20.1.3        23976  online                   1                      0.832
     10.20.1.3    23976                   2  192.168.2.2     23976  online                   1                      0.381
    +---------------+-------+--------------------+---------------+-----------+--------+--------------------+---------------------------+
    2 rows in set (0.01 sec)

                 

How MemSQL is unique

MemSQL is High performance in-memory database with unique features like:

1.  MemSQL in-memory data nodes regularly commit transactions to disk as logs. thus we have advantage to have two copy of dataset per node. One on disk and other on in-memory. hence it delivers results at high speed beating latency that results from reading and writing to disk without any risk of data loss.
2. MemSQL provides lock free transaction so read do not block writes and write do not block read.
3. Provides two tiered architecture. Aggrgators(Master & Child) to accept requests from client and gather information and deliver to them and the nodes which acts as data storage and free for data storage only called leaf.
4. Advantage of keeping database tables in two architecture. Distributed table  that have million of records which is sharded across nodes and reference tables which may have relatively small set of records and can be store on all leaves..

 

MemSQL experience so far

Being DB professional we have a challenge to deliver super performance to our valued customer. High performance could not be delivered with traditional approach so we must need to take modern approach. Here in-memory database technology comes in our mind. this is the only technology which can beat disk base traditional approach.
Now in todays market we have number of in memory options. Choosing one of them is very challenging task as all promise to deliver high performance but as a user we can ensure ourself based on few points
1) Licensing term and fee
2) Maintenance
3) Easy to use.

              I also tested few options available in market but found MemSQL a better option for us. The few reasons are:
1) Simplicity.
2) Lock free DB.
3) In memory with disk storage parallel.

             I am new to this DB but explored quite a bit and like to share all that I explored so far day by day. I hope you will also try by downloading the free subscription available on www.memsql.com .

Good Luck. Cheers.

You can also reach me by email at jitendra_ranjan@yahoo.com and by phone at +91 9810825306