Lippis Report 214: Open Networking Disrupts the Storage Market

nicklippis.jpgThe dialog over open networking often goes something like this: The networking market is fundamentally changing. The established networking vendors will realize lost market share and revenue if they don’t open up. Network switching will become commoditized and offshored to Asia. While these may be true statements, one unknown inter-related trend is evolving, and that’s open networking will disrupt the storage market. The Storage Area Networking, or SAN market, and in particular Fiber Channel, is costly and difficult to configure. Ethernet networking has evolved to the point where early adapters, cloud providers and hyperscale firms are starting to consider transporting storage traffic over the same Ethernet network as user traffic. While 10GbE and 40GbE switching possess the attributes to support storage, open networking’s automation and programmability are the final pieces to the architectural puzzle to enable fully converged data and storage networking. Over the past year, VCs have poured 10s of millions into new IP storage firms to take advantage of this shift in the storage market, including Ceph by Inktank, GlusterFS, Sheepdog, SwiftStack, Scality, Hedvig, Riak CS and others. In this Lippis Report Research Note, we explain the new open networking and storage market that’s emerging.

Attend ONUG Academy and Increase Your Value in the Networking Job Market

Watch the Video

Open networking is all about decoupling networking software and hardware. As this occurs, network hardware becomes less relevant, driving it toward commoditization and lower cost, while software is able to run on different types of hardware that is a network switch or server/x86. With hardware and software decoupled, software can support integration of embedded functionality and external controller applications (via APIs).

The Era of Application Centric Networking

Get the White Paper

Interestingly, storage is going through the same decoupling with costs savings advantages that stem from the decoupling of storage director software from (dedicated) hardware and the replacement of dedicated FC network fabric with more cost effective Ethernet network fabric. Storage director software is evolving from centralized storage control to distributed storage control/cluster solutions. For example, Ceph runs on standard servers.

A Technical Review of Cisco’s Wi-Fi-Based Location Analytics

Get the White Paper

A dedicated storage network fabric can be eliminated, thanks to higher speeds and consistent performance of today’s Ethernet fabric at 10GE and above, which eliminate the once 8/16GE fiber channel speed/bandwidth advantage. Think of it this way: every 12 to 18 months, a new generation of merchant silicon pricing stays mostly the same while forwarding latency is reduced by 50%, jitter or the consistency of low latency is below 10ns, the number of 10GbE and now 40GbE ports double, and power consumption drops.

Unleash the Full Potential of BYOD with Confidence

Get the White Paper

In addition, the new leaf-spine two-tier data center architectures provide non-blocking, fully mesh connectivity, assuring predictable delay and high resiliency similar to fiber channel networks. This cost savings of removing a dedicated FC fabric will be substantial.

There is clearly a trend away from SAN and FCOE due to high cost and static configuration toward scale out IP storage.

10 Gigabit Ethernet: Enabling Storage Networking for Big Data

Get the White Paper

Scale out IP storage appears to be the way to go, and with low to non-blocking leaf spine fabrics, there *should* be enough bandwidth to support. Interesting players in this space include Ceph by Inktank, GlusterFS—which is the upstream OSS project on which Red Hat Storage Server is built, Sheepdog, SwiftStack, Scality, Hedvig, Riak CS, etc. In addition, there are firms in stealth mode, including Avi, CloudGenix, Illumio, etc.

IBM Expands Network Portfolio with SDN VE

Get the White Paper

But it’s not just Ethernet network advances that are driving IP storage into the market that offer favorable advantage over the high cost and static configuration of FC. There is a whole class of important key-value, distributed big data related efforts, such as Hadoop, Cassandra, Dynamo, Riak, Voldemort, Memcached, MongoDB, etc., that are driving IP storage into the market in its own right.

Unified Visibility Fabric Architecture – A New Approach to Visibility

Get the White Paper

You might be asking “what does open networking have to do with IP storage?” The answer is found in automated configuration of compute, storage and networking. As Linux becomes the network operating system of choice, a common set of tools becomes available across administrative domains of applications, servers/virtualization, storage and networking. Tools, like Puppet and Chief for configuration, plus development tools, such as JSON, XMP, Python scripting, REST APIs, etc., enable DevOps to write applications or automation scripts that span all data center assets.

Ballarat Grammar Secures BYOD with HP SDN and Sentinel SDN Security Application

Get the White Paper

In addition, as server virtualization continues to grow so will the requirement of open networking and in particular, virtualized networking that create flows independent of network state. As these flows start to dominate data center networks, server and storage placement become independent, offering data center architects a huge degree of freedom to place servers and storage assets anywhere in the data center. Controllers are setting up flows for IP user data and IP storage.

Open Networking is inadvertently enabling a disruption in the storage market. Smart IT business leaders are reviewing their approach to data center design holistically and finding that there are huge savings potential in open networking and IP storage.

Companies actively engaged in open networking that you can meet with at ONUG include:


One Response to Lippis Report 214: Open Networking Disrupts the Storage Market

  1. Stu Miniman said:

    In my discussions with companies, relative performance is rarely the reason why companies choose between FC and IP. Changing workloads with different requirements are definitely a driver – big data workloads are much less likely to go on a traditional FC SAN. Market forecasts that previously saw a decline in FC now have revenue flat to slight growth over the next few years. If users switch to more service providers and other cloud solutions, it is likely to negatively impact the FC segment. That being said, it is less of a battle of FC vs IP (we are even seeing the hardware supporting both FC and Ethernet today) and more about the solutions that work best for users.