How long is it since we were all asking: “Do we really need Gigabit Ethernet?” Yet, seemingly in no time at all we are all having to seriously consider 10Gigabit Ethernet (10GbE) in a variety of scenarios, in order to avoid LAN bottlenecks re-emerging. The most basic requirement is where a network primarily based around Fast Ethernet connections has moved to Gigabit Ethernet and is making real use of that bandwidth. Suddenly, a Gigabit backbone becomes the bottleneck and a 10GbE replacement is the only way to successfully manage the new traffic levels, says Chris Davies, General Manager of D-Link UK.
Affordability is key
Affordability is a key accelerator here. Thanks to stupendously low costs per port now for Gigabit Ethernet copper connections, people are simply throwing Gigabit into their networks by default. While this, at one point, would have been irrelevant, due to the inability of the client machines, and even the servers, to generate a Gigabit of traffic, thanks to the enormous recent advances in core processor and bus technology, relatively low-end servers are now fully capable of supporting very high bandwidths.
What were single core, single CPU systems are commonly now dual CPU with dual or even quad core architectures. Perhaps more importantly, the Operating Systems (OSs) and applications are now more able to take advantage of these multi-core, multi-processor architectures. And then there is the world of virtualization set to consume even more…
With top research companies such as Dell’Oro Group’s forecasts of major uptake of 10GbE ports over the past 12 months proving correct, despite the current economic climate, the requirement for 10GbE at all key points of the network is becoming ever more magnified. So there is now a valid case for widespread adoption of 10GbE in the core (switches and backbone) as well as at the server and for HPC (High Performance Computing) environments. Moreover, no network manager would typically wish their network to be running at 80-90% capacity, despite LAN switches now proven in running all ports at line speed concurrently.
What is true at the server and in the core uplinks of a network is equally true within the Ethernet switch itself. Whether interconnecting multiple stackable switches or in the backplane of a chassis-based switch, 10GbE is now pretty well mandatory in order to avoid serious bottlenecks emerging and provide truly non-blocking performance (allowing every port to run at line rate). Combining blade servers with Ethernet connectivity in a single chassis emphasises the need for 10GbE all the more.
While the cost of 10GbE compared to Gigabit Ethernet is still high, it is still coming down at a significant rate and is helped by the CX4 copper connectivity option, in addition to fibre connection alternatives. According to a new report from IDC, 10GbE continues to be the fastest growing technology out there, growing 58% Year on Year in Q4 of 2008. IDC believes this increasing demand is due in part to increased scalability and simplified management in the data centre, alongside increased demand for higher bandwidth network infrastructures.
New applications, new requirements
Whereas throwing bandwidth at the problem is never a guaranteed solution on the WAN/Internet due to the inherent latency issues that need to be resolved, in the core of a well-designed LAN, that delay problem is typically far less of an issue. However, the need to guarantee that bandwidth availability is very much a requirement, given the increasing use of real-time applications such as VoIP and video (in its various formats).
In these environments it is essential that spare bandwidth is always available both in the core and at the server, where most of this traffic originates. Using traffic control techniques such as QoS and CoS can prioritise these bandwidth-on-demand type applications, but, in a Gigabit environment there is still a very?excessively?finite amount of bandwidth available in many deployments, so the move to 10GbE becomes inevitable.
There are other significant drivers in the form of new applications and new requirements to enable these applications to be deployed successfully. This even extends to new environments such as Carrier Ethernet. In this scenario where a Provider is looking to provide multiple 100Mbps links into apartment blocks, housing estates or business centres for example, there has to be a way of aggregating this bandwidth and Gigabit clearly is insufficient in this case. Moreover, many of these deployments are looking to support bandwidth hungry applications such as IPTV which guaranteed, nailed-up bandwidth, so 10GbE is a logical play here.
At the same time, many of these customers are looking to a Provider or Hosting company to offer remote backup and storage facilities, downloading?in some cases?terabytes of data. Again, this is a perfect fit with 10GbE at the data centres and all appropriate network nodes between source and destination.
The one certainly in life is that Ethernet keeps evolving. The reason is suitably obvious; as we’ve identified, if a department of users move from Fast Ethernet to Gigabit Ethernet then, quite simply, a Gigabit backbone or switch backplane will become saturated. So that backbone or backplane moves to 10GbE. Equally, as 10GbE becomes more pervasive on server farms, so the 10Gig backbone or backplane becomes the bottleneck and a move to 40Gig is required. For this reason, we have 40GbE around the corner and 100Gig on the way too. However, neither of these are close to being ratified standards yet.
The general view is that the 802.3 standard for 40Gig Ethernet should be ratified sometime in 2010, with 100Gig following maybe 2-3 years down the line. Where 40Gig will make immediate sense is in the backplane of a switch chassis. Here design is always critical in order to avoid ‘blocking’?allowing every port to run at line speed?and the issue will be even more serious at 100Gig. But history shows us that technological challenges in Ethernet are always overcome, so it is very much a case of watch this space.