Backup is a necessary evil, and many would say it’s a little more evil than they’d like.

Backup jobs too often run too long, which may cause other problems such as poor network or application performance during heavy usage hours. Or worse, the backups simply don’t complete at all. And even if they don’t run overtime, a lengthy, slow-running backup job is more likely to be error-prone.

None of this is surprising. What you may find surprising is this: your network may be to blame. As this Search Storage  article points out, “Insufficient IP connectivity is the most common bottleneck in the backup infrastructure. Many customers look at media servers as just another piece of hardware. Most often, they’ll purchase a standard build that may include a dual-ported gigabit NIC. While this may be sufficient for most production servers, it can create serious throughput issues in a backup environment.”

It makes sense that GigE would be the most common backup pipe. It’s cost effective and widely available. The 10GigE alternative remains more expensive and would often be difficult to justify as a dedicated backup network, simply because of the cost and management issues that such a private network would impose.

To Solve the Problem, Look East-West

We could solve the problem with more and fatter pipes, but there is another way. Instead, ask the question, “How can I build a network infrastructure that has the ability to scale up and down in capacity? That is, how can I best accommodate this traffic that is inherently bursty in nature?”

This question – how best to handle bursty server-to-server traffic — turns out to be quite useful because we are increasingly running into applications that generate this kind of traffic.

This server-to-server traffic, sometimes referred to as ‘East-West’ traffic, includes data generated by Vmotion, by clustered servers, and by database clusters. All of these have more east-west traffic than traffic that goes to the core Ethernet switches — often several orders of magnitude more. Cisco will have you believe otherwise, but ask your network engineers whether this is true.

Most existing network architecture is designed to support client-server application traffic in separate pods. This kind of architecture is more focused on top-to-bottom (north-south) flows with a large number of access ports that aggregate into a central core. The amount of oversubscription increases with each layer, making this type of network ill-suited to support east-west traffic.

Virtual I/O Solves the East-West Problem

Xsigo virtual I/O offers an alternative approach. Virtual I/O is fundamentally different from conventional I/O in that all of the I/O resources are consolidated in one location, the I/O Director. Because all of the I/O is co-located, it is possible to move data from one server to another without ever leaving the I/O fabric. And because that fabric can be very fast (up to 40Gb per connection), it is possible to achieve remarkable server-to-server throughput.

You get the ability to provide very high bandwidth to all your servers — backup clients and media servers. And because the I/O is virtual, you can easily provide a dedicated Ethernet infrastructure for each ‘east-west’ application without impacting other production Ethernet traffic. This isolation is achieved by creating dedicated vNICs for ‘backup’ or ‘Vmotion’ or other tasks. These  vNICs terminate on an I/O module which is not used for other purposes.

Xsigo’s solutions allows you to keep all server to server traffic within the I/O Director thereby providing significantly higher total throughput and lower latency. All of this translates to the following:

• A private, dedicated Ethernet infrastructure for server-to-server communication
• Up to 40Gb/s bandwidth provisioned solely for east-west traffic
• Far fewer Ethernet hops and significantly lower latency
• 3x to 5x the performance on backup or Vmotion

One Xsigo customer went from 57 hours to perform a total backup down to 17 hours – an improvement of 70%.

We tend to think of server I/O traffic as being primarily from servers to networks and storage — in other words “north-south.” But it turns out that the “east-west” traffic cannot be ignored. The benefits that virtual I/O offers here can help solve one of IT’s thorniest problems: how to achieve a fast, reliable data backup.