Subscribe / Unsubscribe Enewsletters | Login | Register

Pencil Banner

Four things to consider before upgrading your data center net to 25G

Kevin Deierling, VP of Marketing, Mellanox | Aug. 11, 2016
25G is backward compatible, jacks up performance and improves the efficiency of every workload

This vendor-written tech primer has been edited to eliminate product promotion, but readers should note it will likely favor the submitter’s approach.

Hyperscale public cloud providers and social media giants have already made the jump to 40Gbps Ethernet for their server and storage connectivity for lower total cost of ownership (TCO) and operational efficiency, and now they are migrating to 50 and 100Gbps Ethernet.

Forward thinking enterprises are looking at these hyperscale giants and trying to understand how to achieve Webscale IT efficiencies on an enterprise scale IT budget. Rather than bolting from 10Gbps server connectivity straight to 100Gbps, many are considering 25Gbps as an affordable and less disruptive step that will still provide significant performance improvements.

So, the question is, are you ready for 25G? If not, how do you get ready?

There are four issues to consider when determining your readinenss for 25G:

1.     Which Workloads Benefit from 25G?

Short answer: All of them. Businesses frequently say they don’t run high performance applications and therefore don’t need high speed networks. But this belief fails to take into account the key insight the Webscale folks have learned: High speed networks are not just about performance, they are about efficiency.

As more and more applications become converged, virtualized and distributed, the network is being utilized constantly. This is being further driven by the trend towards hyperconverged infrastructure, which combines compute, storage and virtualization into an integrated platform.

Hyperconverged solutions from Nutanix, Microsoft and VMware all utilize distributed, software-defined storage that relies on the network to move data. Furthermore, snapshots, backups and virtual machine migration is a regular occurrence to achieve load-balancing and high availability.

These ‘background’ tasks consume significant resources and frequently take applications offline or degrade performance while they occur. One solution is simply to over-provision everything; compute, storage and VM instances so that it is possible to provide adequate service level guarantees.

The alternative is to increase network speeds to minimize the impact of background tasks, enabling servers and storage resources to serve applications and users, rather than be consumed simply by moving data. The relationship between higher speed networks and infrastructure efficiency is a key insight that the hyperscale public cloud providers have embraced and has positioned them on the forefront of 25, 50 and 100Gbps network speed adoption.

2.     What is the server, switch and storage infrastructure availability?

As of this month, there are 25G Ethernet solutions from all of the major switch and server vendors available. This is remarkable given that it was just over two years ago (in July of 2014) that the 25G Ethernet Consortium was formed. This initiative was created by a group of companies, including cloud giants Microsoft and Google, to define a specification that enabled interoperable 25G and 50Gbps Ethernet solutions. Since then, multiple suppliers have released silicon and system products that support the standard. Major server vendors including Dell, HPE and Lenovo are now offering 25G, 50G and 100Gbps Ethernet Adapters across a broad range of servers with both traditional stand-up PCI and mezzanine cards available. In addition, a broad range of switch vendors are supporting 25G, 50G and 100Gbps Ethernet speeds.

 

1  2  3  Next Page 

Sign up for CIO Asia eNewsletters.