Shadow IPserver: Fault Tolerance

Only replicated servers can be deployed with confidence

IPserver pairs eliminate a single point of failure.
Automated IP address management simplifies administration, reduces cost, and improves flexibility. Other DHCP implementations have also promised these benefits. However, as a new single point of failure, these same systems compromise network availability.

Through unique and powerful replication technology, Shadow IPservers deliver the benefits of a dynamic IP network without introducing a single point of failure.

Shadow's services collaborate in real time
for reliability and integration.

Backup is not sufficient.
Do you save the memory-resident database from your router to disk? Of course not--the snapshot would become obsolete immediately. For the same reason, a disk backup of dynamic IP address assignments will not guarantee availability.

Frustrated with the futility of database backup, some users have experimented with deployment of multiple servers, each carefully configured to service "half-subnets". The result is complex to administer, and wastes valuable address space.

Instead Shadow IPservers can be deployed in pairs with real-time replication. These pairs constantly communicate to offer the following:

• Replication: Whenever anything in the memory-resident database of one server changes, the other is notified and makes the change. Changes may result from server operations, by client request, or from an administrator management command.
• Load-balancing: While both servers are operational, they provide load balancing for client requests.
• Two-phase synchronization: Before a server grants a client an IP address and configuration, it synchronizes with its peer server, preventing database conflicts. After the assignment, it confirms the change with its peer.
• Uninterrupted service: If one server goes offline, the other server continues to service all clients.
• Automatic re-synchronization: When a server comes back online, it automatically synchronizes its database with the peer.

Enable responsive, centralized configuration using short lease times.
If, for any reason, a DHCP server becomes unavailable, it impacts not only new clients who cannot attach to the network, but existing clients currently using the network. When an address lease expires, open network connections are simply abandoned and associated network applications will quit working, or crash.

Alternative automated IP management implementations limit the risk of disruptions by configuring very long lease times (a week to a month) as a hedge against server failure.

But long lease times reduce the value of automated systems. Long lease times mean clients are slow to learn about changes in the network environment, paralyzing the network and the administrator's ability to meet the needs of the business. Long lease times limit the ability to reuse scarce addresses, and may not be feasible for implementation of VLANs or policy-based management.

Perform all system administration online without downtime.
Because Shadow IPservers are fully replicated as peers, system administration does not require scheduled downtime. This also means that servers can be updated, moved, or replaced on-line as necessary during normal working hours with no disruption to services.