Sunday, May 11, 2008

Gigabit Ethernet Applications-Is It A Fit For Your Organization?

In general 10 Gigabit Ethernet links are deployed in parts of an enterprise or service provider network where large numbers of Gigabit Ethernet links are being aggregated. This most commonly occurs in the network core, but as gigabit-to-the desktop deployments continue to grow the need for 10 Gigabit Ethernet uplinks from the network edge will emerge. Following are some common uses or applications for 10 Gigabit Ethernet.

Enterprise

As all enterprise applications converge on the common Ethernet backbone, the network must be able to gracefully support a new array of bandwidth hungry applications such as storage, backups and video. To effectively deal with this onslaught of bandwidth demand, forward-thinking network architects are designing enterprise networks with 10 Gigabit Ethernet to meet current and future needs.

Starting from the edge of the network, as gigabit-to-the-desktop deployments continue to grow, the need for 10 gigabit uplinks from a wiring closet switch to handle the performance demands of power users will emerge. Backbone links to entire floors or buildings may run over optical fibers at gigabit rates, while drops to desktop workstations may need to be no faster than 100 Mbps. Naturally, with this upsurge in bandwidth, demand 10 Gigabit Ethernet will be used as a high-speed interconnection between multiple buildings. These buildings could be in close proximity on a self-contained "campus", or could be many miles apart and be connected by dark fiber provided by a local service provider or municipality. Due to the limited distance of 10 Gigabit Ethernet on multimode fiber (MMF), singlemode fiber (SMF) must be deployed or leased to support 10 gigabit building interconnects.

In an enterprise data center, 10 Gigabit Ethernet can be used in a variety of applications such as cluster computing, server attachment and storage interconnect.

High Performance Cluster Computing (HPCC)

Server clusters are a group of tens, hundreds or even thousands of relatively inexpensive (e.g. 1U Linux servers) computers (referred to as nodes) connected in parallel to cooperatively solve large, complex problems. Cluster computing has become a mainstream technological tool for research, financial modeling, digital image rendering and scientific applications. The connection between the various computers in the cluster is typically an Ethernet switch.

Gigabit Ethernet Servers and NAS

As servers and Network Attached Storage (NAS) devices are attached to the edge of the network at 10 gigabit speeds, the network core will have to scale proportionally by link aggregating multiple 10 Gigabit Ethernet links or eventually with 40 Gigabit Ethernet. Network architecture must take into account meeting current Gigabit Ethernet and 10 Gigabit Ethernet server connectivity needs and scale upwards as the need for greater bandwidth inevitably emerges.

Storage Interconnect (iSCSI)

Fibre Channel has been the protocol of choice for storage area networking, however, with the emergence of the IETF (Internet Engineering Task Force) iSCSI protocol, Fibre Channels dominance is being challenged by IP/Ethernet. Simply put, iSCSI is an IP-based storage networking standard that facilitates data transfers by carrying SCSI (a protocol commonly used for communication between storage devices and computers) commands over IP networks. As iSCSI continues to gain in popularity, 10 Gigabit Ethernet links are the logical choice to carry the enormous volumes of data that traverse large corporate networks.

Given the extreme complex nature of designing an appropriate Gigabit Ethernet network architecture....and even deciding if it's right for your organization....it's strongly suggested that you not dump this whole animal on your IT staff. At a minimum....make use of free technical consultation for the bandwidth sourcing requirements from DS3-Bandwidth.com.

Michael is the owner of FreedomFire Communications....including DS3-Bandwidth.com and Business-VoIP-Solution.com Michael also authors Broadband Nation where you're always welcome to drop in and catch up on the latest BroadBand news, tips, insights, and ramblings for the masses.

Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx

When Is Ethernet A Good Choice For A Business Network Infrastructure

Many businesses today are showing more and more interest in choosing ethernet applications as part of their overall network infrastructure. However, too often they do so without arming themselves with enough information to make a smart business decision. Word of warning .... get smart first.

Ethernet is a multiple access technology where the entire bandwith is available to each user on a contention basis. T1/DS3 on the other hand is a (time-division) multiplexing technology. Where in fractionilized form it will allow users continual access to the their own individual 56K/64K timeslot.

Now, while we can discuss grouping timeslots or making the entire bandwidth available, there is one aspect missing in the T1/DS3 argument and that is contention - simply how do we allocate the bandwidth to multiple users. Ethernet employs this by design. But I would expect that T1/DS3 needs an additional resource to implement this .... especially for a voice application aspect. Now, I guess that there is an argument for utilizing a real TDMA technology. But then we come down to cost. Ethernet is pretty damned cheap and can offer flexibility of design/installation. Where traditional TDMA technology relies on pretty precise timing across multiple hosts within the network and therefore becomes increasingly expensive.

Today businesses are moving to Ethernet based Internet networks because of:

- Speeds Available: For example Time Warner Telecom offers 2 Mbps - GigE.

- Scalability: very easy to upgrade speeds and can be deployed via copper or fiber optics: T1, DS3 or OC3 - OC192.

- Hardware Costs: less expensive and easier to maintain. A 10/100 Ethernet Card in any Layer2 device (i.e. Firewall) or Layer3 device (i.e. Router) is all you need to terminate your service.

- Network Simplicity: instead of having a router and firewall, you only need one piece of hardware and companies are using Firewalls that have routing capabilities (which eliminates the purchase and maintenance of a router). This helps out your IT budget, equipment maintenance (your LAN is already Ethernet) and you have less points of failure on your network (the more hardware devices you have, the more points of failure in your network).

- Fiber Optics: ISP's would rather deploy their fiber optics than lease DS3 local loops from RBOC's or ILEC's. Carriers are expanding their fiber optic footprints and more ISP's are offering fiber optics to deploy faster and better quality Internet for businesses. Thus the ethernet connection.

Overall Ethernet is great for businesses on three counts:

1) Price Point - the cost of Ethernet devices and interfaces is significantly less than the cost for the others you mentioned because, among other things, market penetration of Ethernet means economies of scale come into play, driving the cost down. In addition, Ethernet is usually a non-tariffed service, so market competition allows for the price to come down as well (as compared to tariffed services supplied by the telcos).

2) More engineers understand it - it's a lot easier to troubleshoot and maintain because a significantly higher number of engineers/admins understand and are comfortable with Ethernet. Having an optical engineer or telco engineer on staff will cost you much more than an Ethernet "engineer".

3) Ubiquity - it looks like the big players are driving the market (or are being pulled by the market) towards Ethernet. That means that most technological innovation is directed towards Ethernet, not T-1's, DS-3's etc.

With that said, I don't think Ethernet is going to supplant optical offerings, particularly DWDM, in the long-haul market. Timing and error correction are too critical to long-haul transport and Ethernet doesn't really have either. So, I see Ethernet as an optimal last-mile technology that is carried over optical services.

One caveat, if your business model needs to support legacy telco applications (non VOIP) along with outside bandwidth, then channelized T services may be more appropriate. That is, if you want 15 channels of a T-1 for voice and 9 channels for Data that may make more sense than having two separate connections, one for voice and one for data.

I doubt that anyone would disagree with Ethernet being the absolute best protocol for a businesses internal LAN. But I have been seeing more and more ISP's trying to deliver their data already Ethernet encapsulated. Rather than provide standard conditioning over a DSL, T1, DS3 ..... they want to transport the data already Ethernet encapsulated. This is usually done to allow them to offer additional services such as a firewall at their NOC, rather than your site. But you give up a lot by letting them convert your data to Ethernet packets at their NOC.

First you give up bandwidth. They still have to transport your data over their pipes to your location using whatever form of encapsulation needed by that transport (DSL, T1, DS3, etc.). But there are the additional layers needed for Ethernet which is now added to the layers for the transport. In other words, the data from your files will have more overhead or layers for the same amount of data. A given file size will take longer because of the additional data required for the capsules.

You also lose trouble shooting ability. If a business is connected using standard transport, an ISP can ping right to the router in their facilities. If they are forced into a PPPoE delivery, the ISP cannot check their connection between the NOC and their location. Often it is a last mile (wires to the building) problem and there is no way to check that.

I could go further, but the answer is that for a business LAN, Ethernet is the way. But it is NOT for your internet connectivity to your ISP.

Now here's a few other tidbits often overlooked by a business when considering Ethernet as the backbone of their network infrastructure......

Ethernet is not cheaper than any other layer 2 protocol, also not in WAN bandwidth. Order a 100 Mbps from your friendly neighborhood provider and ask them specifically that you want a Sustainable Information Rate of 100%. You will be shocked to learn that it is nearly as expensive as an OC3 bandwidth. It becomes cheap because Ethernet is commonly overbooked without providers telling you that.

Yes Ethernet has it's advantages. But definitely not the advantage that you no longer need Customer Premises Equipment. On native Ethernet you can use layer 2 (VPLS). In which case you will have to maintain the switches on the WAN side. Or you can have layer 3 VLAN services. In which case your provider most likely will manage the routers.

Be aware that it is the VLAN's that are interesting. They are somewhat comparable to MPLS Classes of Service. Which means you can have a managed VLAN for different applications, with different service parameters. A provider that sets up his Ethernet backbone infrastructure correctly, would be able to provide high quality IP telephony over a TLAN with decent Quality of Service parameters (especially low jitter).

Note that if you were to purchase a fully managed MPLS VPN from most any company anywhere in the world, Ethernet would not bring you much more flexibility. But if you are going to need a lot more bandwidth with bursty character, you should target an Ethernet layer 3 managed network. No matter when a 100 Mbps Ethernet link is overbooked, most often you will still get more bandwidth and less congestion than on the SDH/SONET bandwidth parameter costing the same.

A key point is that at present native Ethernet coverage is still a myth in most parts of the world. Most often some electrical 155Mbps is simply squeezed down to 100 Mbps, because there simply isn't fiber at a location yet. 1Gbps or 10 Gbps is in that case not available or unbelievably expensive.

On Layer 3 this squeezing of SONET bandwidth does not pose a disadvantage. Because Layer 2 becomes near to trivial anyway, and a hybrid network is fine. You can easily hook up Classes of Service of MPLS with VLAN's. On Layer 2, there are not many providers that can provide a VPLS over any hybrid infrastructure network.

Now here's the bottom line (pun intended).......

In terms of quality and price-performance ratio, the supplier and your specific applications are much more important than the technology. However, currently for Internet applications ..... and purely because of the overbooked price difference of large bandwidths ..... Ethernet is preferable for most businesses.

Michael is the owner of FreedomFire Communications....including DS3-Bandwidth.com and Business-VoIP-Solution.com. Michael also authors Broadband Nation where you're always welcome to drop in and catch up on the latest BroadBand news, tips, insights, and ramblings for the masses.

Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx
Streamyx