Fault tolerance

Fault tolerance SearchSearch
Author Message
Alex Kaiser
Frequent Contributor
Username: Alex_k

Post Number: 85
Registered: 07-2006
Posted on Wednesday, February 24, 2010 - 06:06 pm:   

Hi!

In topic http://support.nowsms.com/discus/messages/53/23981.html, like it was sad - the fault tolerant server stores user data balances but fault tolerant (data protected) doesn’t mean high availability. Server just can stop recently because hardware or software problems – all cluster become unusable! Indeed, you won’t loose any data because of disk arrays or something but just imagine approximate recovery time for server 2000km far from you, somewhere at ISP datacenter. I strongly believe that users billing data is most important NowSMS’s subsystem, so other data possible can be stored locally.
My suggestion is to create some caching mechanism at SMSGWs. Let me explain - 3 servers STOR (redundant with billing data), SMS (SMS Router). SMS has UserDir=\STOR\UserData and doing all transactions over network. If STOR has problems (ex. NIC failure) then sending stops. If after each transaction SMS create local temp folder (ex. <NowSMS>\Transactions\<UserLogin>) and overwrite balance file (<yyyymmdd.ctr>) with RouteCharge and Balance if transaction was successful and if it’s not just use last line to calculate balance:

20100224.ctr
1;10256
2;10254
3.5;10250.5
0.5;10250

After when connection with STOR will be restored SMS can easily synchronize balances.

Do you like my idea?

Regards,
Alex K.