Enterprises often fail to adequately plan and provision for recovery from disasters resulting in large downtimes and related disruption to business. It’s important to understand that even servers under warranty can fail and data loss can occur (even if the vendor adheres to the response and resolution times committed to bring the server alive).
|Practice||Technique||Description||Time to restore services|
|Basic Availability||Backups||Ensure backups of all servers are taken to a separate PC/server in the network, onto a secondary device (tape, USB device) and a copy kept offsite with rotations done at regular intervals. The recovery procedure is to procure hardware, install the OS, install MCS and restore the backup.||8-10 hours|
|Basic Availability||Standby server||Maintain a hot standby server with OS and Mithi Connect Xf installed and ready to be deployed in case of a failure. This will have to be kept up to date with all upgrades, services packs and hot fixes as they are deployed on the Primary. Keeping this handy will save time while getting a secondary to replace a primary. It will require the data to be restored from the backups to take it live.||2-3 hours|
|Enhanced Availability||Disaster recovery solution||Deploy the Disaster recovery solution for Mithi Connect Xf. This keeps the secondary server up to date in real time and provides a simple switchover to the secondary in case of a failure on the primary. The secondary can also be kept at a remote location.||15 mins|
|High availability||Load balanced setup||Deploy the HA solution for Mithi Connect Xf. It comprises redundant compute only servers attached to a shared storage, accessed via a clustered set of load balancers. If a server is to fail or even get overloaded, the load balancer will take it off line and continue distributing the load to the remaining servers.||Hands free instant|
Enhanced availability using an Active-Passive setup
Disaster recovery using a hot standby server
An organization where mailing infrastructure is the key for business continuity and would like to achieve availability with minimal cost. One of the key concerns of IT Managers is to how to best quickly recover from a mail server failure.
Mithi Connect Xf can be setup as a disaster recovery or a hot standby server for a mail server hosted on Mithi Connect Xf.
The Email setup ensures that in case the primary mail server is not functional or not available due to a hardware or network failure, the secondary server takes over automatically.
The hot standby server will have the account data and the mail store from the primary mail server. When setup as a disaster recovery server to a mail server hosted on Mithi Connect Xf, the process of updating of the hot standby server is automatic.
- Real time synchronization of data keeping the hot standby server updated.
- Minimal server downtime as users can be switched to the hot standby server in case of a primary server outage.
- Immediate recovery from the failure as primary server can use configuration and mail data backup from the hot standby server for reverse synchronization.
- Low cost as no high-end software or hardware other than Connect Xf Email Server is required
As a hot standby server for Mithi Connect Xf
The solution uses a low level block synchronization technology (DRBD) which ensures on the fly replication of blocks. In case of an outage, it is possible that a few blocks may not have got properly synchronized leading to a localized data loss. The hot standby server capacity is not utilized during normal functioning since it is a passive entity.
High availability using an Active-Active setup
Load-balanced setup using a shared network storage with auto-fail over
An organization where the messaging infrastructure is the key for business continuation and would like to achieve near-zero downtime with auto fail-over in case of a server failure.
Mithi proposes deploying two or more MCS servers, which handle the SMTP, HTTP, POP, IMAP and LDAP traffic. These servers use a common NAS (network attached storage) for the mail store, such that they can simultaneously read or write from the mail store for any user.
Clustered load balancers will route all IP traffic from the clients to any of the servers that is available and responding well. The load balancers ensure that there is no downtime and the fail-over to available servers is automatic. To the end-user, all the MCS servers appear as one server.
The servers can be of different configurations, however Mithi advises that the servers be of identical configurations.
The load balancers take special care of:
Load balancers can be built on commodity hardware using open source components.
Addition of servers will enhance the performance of the system as more processing power will be available. Separating the incoming and outgoing mail traffic will ensure faster delivery to and from the Internet.Scalable
The architecture proposes deploying a farm of servers connected to a load balancer and have the storage on a common storage device such as a NAS. This architecture ensures that on addition of users or an increase in the mail traffic can be handled by addition of mail servers to the farm.Highly available
The architecture ensures minimum downtime if any of the servers develops a problem, For example if one of the mail servers fail, then the others can assume the responsibility of handling the traffic, till such time as the other server is out of commission.Mail archiving
The new architecture can be configured to archive all traffic to SATA disks on the NAS.Optimal use of infrastructure
This architecture ensures that all the servers are used since they all are active. The NAS storage solution can be shared for other enterprise applications to help consolidate the storage requirements of an organization and also exploit the strong native features of a NAS for backup and data redundancy.
Resources : Preventing a Disaster
Quiet but persistent and firm, she assists Sunil to manage the Services division.