10 GbE: the next big thing
10 GbE over copper up to 100 metres (802.3 an) is expected
to be ratified by mid-2006. So is it the next iSCSI in the making? Russell Julius,
Technical Director, Systimax Solutions, Asia Pacific, gives his take in an interview
with Anil Patrick R
What are the verticals that you will see adopting 10 GbE
We see the pharma, research & development, and financial segments as the
first adopters of 10 GbE over copper. This will typically be accompanied by
organisations with higher bandwidth demands and the bottom-line wherewithal
to pay for the technology. These are the initial typical adoption areas in addition
to data centres and the use of 10 GbE in risers.
However, surprisingly all the sites that I have been involved with on the Cat
6a cabling side have chosen 10 GbE compatible cabling across the network. It
is as if we are going to do this, we are going to put this everywhere.
So from what you have observed so far, it is not just part
of the network that is getting 10 GbE treatment, but the entire network?
First, they are not using 10 GbE on copper yet, simply because the hardware
is not available. However, when the time comes to wire up the network, it is
not as if only the finance or the R&D people need it. The approach is that
10 GbE has to be all across the enterprise.
This is despite the fact that Cat 6a is more expensive
than Cat 6 at present?
When you look at the scale of things, you might be paying extra for that cabling,
but it is not that big a differential either. Compared to the rest of the non-IT
infrastructure used across the organisation, it is hardly a huge or bad investment.
What is the price differential of Cat 6a compared to Cat
6 or Cat 5e?
I cannot give you an absolute figure since that varies from place to place.
It depends on the cost of the labour. In the higher-cost labour markets, the
percentage difference will be lower than that in lower-cost labour markets.
So what about India? Will it be cheaper than other countries?
Certainly, the labour costs in India are cheaper than a place like Australia.
The cost difference between Cat 6a and Cat 6 is lower in India than the cost
in, say New Zealand. This is because the labour component cost is higher in
New Zealand than in India, but the cost of material is almost the same.
The actual throughput in the case of Gigabit Ethernet is
much less than one Gbps. So how much will the 10 GbE be able to pull off in
terms of transmission speeds?
Ethernet as a protocol is not exactly the most efficient thing in the world.
The efficiency rates in the case of 10 GbE over copper would essentially be
the same due to the common protocol being used.
Therefore, the truth is that gigabit is not 100 percent efficient, nor is it
more than 98 percent. The actual figures are more likely to be in the range
of 60 to 70 percent efficiency. Therefore, we can expect to see the same efficiency
rate in 10 GbE as well.
So that will be in the range of
Well, at a 70 percent efficiency rate, actual throughput will be in the range
of more like 7 Gbps. That is the reality of Ethernet technology.
Talking about cabling technology, what challenges did you
face in designing 10 GbE over copper?
There are many challenges. There is the internal performance challenge of the
cable and connectors, but people talk most about the external challenges. This
is the cross talk between two cables.
As you know, internally the cables consist of four pairs and each pair is twisted
in a different way. This is to prevent them from cross talking with each other.
Conventionally, when you lay two cables side by side, there might be chances
of like pairs being close to another like pair and cross talking.
Now, at the high frequencies that we are talking about, we can actually take
leakage from one cable, one channel to another channel because of those like
pairs. That is why we use a unique finned arrangement in the jacket for extra
physical spacing between the cables. You do not see that in Cat 6 cables because
the frequency is limited to 250 MHz as opposed to 500 MHz in Cat 6a.
While cross talk was a problem in gigabit Ethernet,
it was relatively easy to overcome. However, in 10 GbE cross talk is a
bigger problem. Compensation within the connector itself is different
due to design for performance
While cross talk was a problem in gigabit Ethernet, it was
relatively easy to overcome. However, in 10 GbE cross talk is a bigger problem.
Compensation within the connector itself is different due to design for performance.
In terms of connector terminations, there is not much difference between the
termination processes. In terms of what an installer does to terminate a connector,
it is not very different from Cat 6.
As a standard, how do you see the roadmap of 10 GbE?
ISO has recently published a channel standard. The permanent link and component
requirements will follow on. I expect the channel requirements to be published
as part of the ISO/IEC 11801 standards. That is the key international standard
for commercial Cat 6 cabling.
So will 10 GbE over copper provide competition to the use
of fibre over short distances?
Absolutely, that is why we see organisations use 10 GbE in data centres and
wanting to use more of it. It is a cost-effective way of doing it. However,
there is a limit of 100 metres. The copper channel has been clearly defined
as up to 100 metres when application independent cabling first came in many
years ago. That is like the holy grail of cabling.
Once you go over 100 metres, you have to change over to fibre. Then you need
fibre to go to longer distances. Below the 100-metre range, copper definitely
provides a price advantage over fibre.
As of now, how many networking vendors have 10 GbE products?
I am not aware of any vendors with 10 GbE products in the marketplace.