Problem solve Get help with specific problems with your technologies, process and projects.

Multipath I/O and failover: The differences

Storage expert Rick Cook discusses the major differences between failover and multipath I/O and clears up common misconceptions about each.

What you will learn from this tip: Storage expert Rick Cook discusses the major differences between failover and multipath I/O and clears up common misconceptions about each.

Both multipath I/O and failover rely on multiple redundant links between the server and storage. However, they use the links differently and often for different purposes.

The situation arises because of the fundamental mismatch between modern communications channels and server and storage speeds. Servers and storage devices can run a lot faster than a single connection between them can support. Multipathing handles the mismatch by breaking the data flow into multiple streams and sending each one over a separate channel.

Failover uses one or more alternate links between the storage and server to provide redundancy in case the primary link fails. Multipath systems always have multiple links available. Failover only requires an alternate link when the main link goes down and the alternate link's logical connection may not exist except in case of failure.

What confuses the issue is that multipathing and failover are easy to combine and often are combined. Vendors offering multipathing can easily add failover features, and many of them do. Going the other way -- adding multipathing to failover -- isn't quite as easy, but it is still a popular added feature for storage systems with failover.

It is important for storage administrators to understand that having one doesn't mean you necessarily have the other. Adding multipathing to a system with failover or vice-versa is probably going to take some additional installation and configuration, and it may require additional hardware and software as well.

In storage, some degree of failover is inherent in multipath installations with load balancing, as the remaining paths will pick up the load from a failed path. However, the opposite isn't necessarily true. Pure failover only activates when the primary path between the storage and server becomes unavailable. Further, failover provides additional features, such as failure alarming, that aren't necessarily part of a multipath installation.

There can be some architectural differences as well. For one thing, a truly reliable failover installation requires complete redundancy between the storage and server. Multipath I/O paths may share components, especially SAN switches, although multipath installations typically use more than one HBA and physical connection between the server and storage.

Microsoft outlines its version of multipath I/O and discusses the relationship to failover (and clustering) here.

Do you know…

How to use clusters to ensure failover?

About the author: Rick Cook has been writing about mass storage since the days when the term meant an 80 K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last 20 years, he has been a freelance writer specializing in storage and other computer issues.

Dig Deeper on Data storage strategy