Welcome to the continuation of our open test program aimed at validating the performance, service scalability and power efficiency of 100Gigabit Ethernet (100GbE) router interfaces. As more and more vendors are beginning to deploy the first generation of 100GbE interfaces for their flagship routers, European Advanced Networking Test Center AG (EANTC) and Light Reading have set up a public test program aimed to provide operators with an unbiased view of these new interfaces.
As was the case in our most recent test, we took an up-to-date mix of carrier network requirements and created a test plan that we intended to apply to all vendors participating in the program. Vendors also had the opportunity to define two of their own tests in addition, at our approval.
With the help of testing vendor Ixia, then performed agreed-upon tests at each vendor’s lab. The results will be presented in a series of articles right here on Light Reading -- one for each vendor who goes through the tests.
This report will recap the results for Alcatel-Lucent (NYSE: ALU), our latest test participant. Alcatel-Lucent's 7750 SR-12 has been around for some time and the company showed that adding 100GbE interfaces doesn't change much about the features operators are already familiar with. Alcatel-Lucent sometimes has their own way of doing things, but as you'll read they ably took on the challenge of this test.
Testing Info
We worked closely with Ixia Communications to ensure that
our scalability goals were reached while meeting our No. 1
goal of testing services realistically. It's also worth
reminding those who don’t use these tools day to day, that
long passed are the days of so-called "packet blasting."
Thankfully, IxNetwork -- the software we used for all of the
tests -- allowed us to emulate more realistic scenarios.
MPLS services testing requires intelligence in the tester to
represent a virtual network with hundreds of nodes.
IxNetwork calculates the resulting traffic to be sent to the
device under test across the directly attached interfaces --
including signaling and routing protocol data as well as
emulated customer data frames. This way, a reasonably
complex environment with hundreds of VPNs and tens of
thousands of subscribers can be emulated in a representative
and reproducible way. For the 100Gigabit Ethernet test
interface we used Ixia's K2 HSE100GETSP1. The software used
was IxNetwork version 5.70.120.14, IxOS version 6.00.700.3.
About EANTC
The European Advanced Networking Test Center (EANTC) is an independent test lab founded in 1991 and based in Berlin, conducting vendor-neutral proof of concept and acceptance tests for service providers, governments and large enterprises. EANTC has been testing MPLS routers, measuring performance and interoperability, for nearly a decade at the request of industry publications and service providers.
EANTC’s role in this program was to define the test topics in detail, communicate with the vendors, coordinate with the test equipment vendor (Ixia) and conduct the tests at the vendors’ locations. EANTC engineered, then extensively documented the results. Vendors participating in the campaign had to submit their products to a rigorous test in a controlled environment contractually defined. For this independent test, EANTC exclusively reported to Light Reading. The vendors participating in the test did not review the individual reports before their release. Each vendor had a right to veto publication of the test results as a whole, but not to veto individual test cases.
— Carsten Rossenhövel is Managing Director of the European Advanced Networking Test Center AG (EANTC), an
independent test lab in Berlin. EANTC offers vendor-neutral network test
services for manufacturers, service providers, governments and large
enterprises.
Carsten heads EANTC's manufacturer testing and certification group and
interoperability test events. He has over 20 years of experience in
data networks and testing.
Jonathan Morin, EANTC, managed the project, worked with vendors
and co-authored the article.
The article stated that the 100GigE interface was an SR-10. To my knowledge, this interface is a parallel interface with multiple 10G lanes. Aggregated to 100G.
Question: is this true? That the "100G" side of the tests were actually multiple 10G physical lanes?
If so, how is this different than the other side of the test where it was acknowledged that the links were indeed multiple 10G lanes.
I realize SR-10 was used due to availability of interfaces from ALU suppliers. And that the statement that the box also supports CFP LR4 interface, but that interface was not used in the tests. Even though the SR10 is a "single" interface, it is actually multiple 10G lanes each over it's own multimode fiber.
So, some clarification would be appreciated. It seems to me that this entire test was basically multiple 10G lanes talking to multiple 10G lanes with one side aggregating the 10G lanes to make a "100G" port. And the other side treating the 10G lanes as independent oversubscribed ports.
Still an impressive achievement from the box and the tests. However, it seems clear the electrical IO was all still running at 10G.
Actually, I think the SR10 standard is also 20 strands of fiber. Not 2. As I read it anyway. Willing to be corrected if I mis read the spec.
2 strands would require that this is single mode, and WDM muxed with 10 lambdas and a demux at the end. if it is indeed WDM muxed then I mis-read the spec for SR10 and my mistake.
Yes, 40G and 100G all use 10G lanes. The difference comes in the fact that you have 100G over a single connection; so two strands of fiber were used; one for tx and one for rx. On the ten 10G links, 20 fiber strands were used; 10 for tx and 10 for rx.
In the real world, if you needed 100G between A and Z, would you rather use 2 strands or 20? Sure 20 provides some redundancy over a failure, but you could also use 4 strands and get dual 100G rather than just 100G by using twenty strands. Another issue that you run into, most equipment does have a limit to the number of aggregated pairs you can have as well as the number of ports in the aggregation. You are usually limited to 8 per group, so in a single group the most you could get is 80G. So in order to get 100G using individual 10G interfaces, you would have 10 totally separate paths. The 100G is making it all look like a single interface as well as you are not using a lag to accomplish it.
Actually, the interfaces used were LR-10. LR-10 are 10 lanes of 10Gig DWDMed into a single pair of fibers. Also, the electrical interface for the CFP is ALWAYS 10x10 Serdes. LR-4 optics using 4x25 Gbps lanes with DWDM muxing require a gearbox chip to convert the 10x10 into 4x25. Keep in mind though that these 10x10lanes (on the CFP-baseboard) interface are bit-level muxed and have no view on packets whatsoever. All interfaces, be them LR4, LR10 or SR10 allow the system to carry a single 100Gbps flow. There are no changes in the packet processor or data plane depending on the optics plugged into the card.
100GBASE-SR10 is an IEEE standard defined interface for 100 Gigabit Ethernet. I don't believe that the full clause is accessible for free online, but I'll point you to the presentation that was the basis for the eventual detailed specification for this PMD interface.
You are correct that the 100GE packet is striped across a physical bus of 10 optical lanes, however, the key point is that this is a 100GE packet. The IEEE standardized the SR10 as a low cost 100GE variant for <100m connectivity.
Note I do not work for ALU, nor do I have any knowledge of their product -- I'm just chiming in on the IEEE standard.
There are only three types of CFPs available for 7750 right now: LR4 10km, OTU4 LR10 10km and LR10 4/10km (4km OTU4 or 10km 100GE). SR10, for sure, is typo.
Thanks for the clarification regards LR-10. I am very aware of the LR-10 spec. If the article had stated that, I would never have made my post. The article stated that it was an SR10 interface.
Hence the comments.
I am also quite aware of the 10x10 SerDes on CFP and the gearbox requirement for CFP to use 25G lanes. I have been very involved with 10G, 40G, 100G (including 4x25 and all the various reaches in the standards and proposed standards) applications for many years.
And I get it regards the packet processor, the data plane and bit level muxing.
if the original article, or the editor or the ALU company representative who should have reviewed it before publication for accuracy had caught it, and the article had not incorrectly stated SR10, this discussion would never have happened.
As correctly mentioned by a few others, the CFP modules used in this test were in fact LR10 variants (SR10 is a typo). Of course the LR4 variant is also supported and shipping. In either case for LR4 or LR10 there is an embedded WDM mux/demux function such that the 100G interface is served over only 2 fibers (TX & RX) and not 20.
Both cards that ALU has; single and dual port 100GigE all handle it with two fibers per port.
On the Ixia side going into the SR-12, they had 10 physical ports and each were not dependent on the the others; so no lag was used as that maxes out at 8 ports per lag group. Each port used 2 fibers; so 20 in total. On the 100GigE interface, it was two fibers going into a single port. What they were showing with the test, the IMM could handle 100G of data. So 100G into a single slot. The FP3 that ALU is using can handle 400G of throughput. So 4 port 100G cards will be available in the future.
With the card being run by the FP3, it won't matter if the lanes are 10G or 25G, the same physical silicon is handling it. So the test of 4x25 will produce the same result.
They should have also mentioned what the forwarding rates were for the Multicast groups.
The "LR10" is not an IEEE standard. It is a multi-source agreement based upon a proposal from Santur. There are many pros & cons associated with it. For a list of the pros, see http://www.10x10msa.org
To save this item to your list of favorite Light Reading content so you can find it later in your Profile page, click the "Save It" button next to the item.
If you found this interesting or useful, please use the links to the services below to share it with other readers. You will need a free account with each service to share an item via that service.
(14) | 
Digg
Del.icio.us
Reddit
Tweet This