x
Optical/IP

Testing Cisco's Mobile Core, Data Center & Business Services

Summary: Cisco demonstrated the transfer of 20 Terabytes of data, using maximum jumbo-sized Ethernet frames, at line rate across Nexus 7000 and CRS-3 (the latter equipped with 100-Gigabit Ethernet) in 26 minutes and 40 seconds.

Cisco’s Data Center Interconnect (DCI) System is a solution used to extend transparent Ethernet segments (subnets) beyond the traditional boundaries of a single-site data center. By stretching the network space across two or more data centers, the DCI solution is able to facilitate the enterprise’s server virtualization strategy, support high performance, and provide nonstop access to critical business applications.

In this test, Cisco asked us to verify that Cisco’s data-center-to-data-center interconnect, accomplished by connecting two Nexus 7000s over a pair of 100-Gbit/s attached CRS-3 routers, could transfer very large data sets while enjoying the full benefits of the newly introduced 100GbE interfaces of the CRS-3. Cisco used a back-of-the-envelope calculation to figure out that the library of congress (20 Terabytes of data as the common belief goes) could be transferred, theoretically, in 1,600 seconds from one data center to another. Cisco was so sure of the ability of the solution that they challenged us to execute exactly this test. (The main 100-Gigabit Ethernet tests on the CRS-3 are documented in the first part of the article.)

We attached ten 10-Gbit/s tester interfaces to each one of the two Nexus 7000s and started sending the largest possible Ethernet frame sizes (over this medium: 9216 bytes) from one end to the next, clocking our transfer at 1,600 seconds (26.66 minutes). The question was simply: Had 20 Terabytes been successfully transferred at the end of the test interval?

The test setup emulated 10 file servers on each of the two data centers. Half of them were used to source the traffic and the other half were used to receive the data transfer - much like off-site backups that happen regularly across geographically dispersed data centers.

As a result, the test confirmed Cisco’s claim: The full amount of data was transferred at line rate within the time calculated in advance.

Page 14: Conclusion

Previous Page
13 of 14
Next Page
Page 1 / 2   >   >>
srikrishnak 12/5/2012 | 4:24:27 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Great continuation. Used to get much more from LR, so still waiting for PDF :)

DCITDave 12/5/2012 | 4:24:21 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

We'll have the PDFs available soon. Want to make sure all corrections and copy editing are complete with both reports first. Hang in there.

JeddChen 12/5/2012 | 4:24:21 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Informative Reports. waiting for PDF, too.

bigpicture 12/5/2012 | 4:24:17 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

I have a question about the throughput vs. the DPI tests.  In the throughput test you stated you had 60,000 make / break sessions activating at 1,000 sessions per second.  In the DPI test you stated the rate was "reduced" to 5.000 sessions per second.  


I'm not sure how you "reduced" from 1,000 to 5,000.  Is one of the numbers a typo or am I just confused?


Thanks,


BP

tohhwee72 12/5/2012 | 4:24:17 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Possible to share more detail on the firewall performance, e.g.


1. The packet size used for the test, i.e. large or small packet?


2. Traffic profile, e.g. % of http, % of ftp etc


3. How is the NAT being done? i.e. how many public IP address are used to get 1 mil concurrent session?


4. What was the throughput, CPU load etc?

cross 12/5/2012 | 4:24:16 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Hi Bigpicture,


I thought I was confused, but it was only a typo :-)


Thanks for pointing out the issue.  We will fix it.  In short:

<ul>
<li>GGSN test with DPI: 18,000 attachments per second</li>
<li>GGSN test without DPI: 18,000 attachments per second</li>
<li>EPC test: 5,000 attachments per second.</li>
</ul>

Once all sessions were established, the make/break attachment rate was 1,000 new sessions plus 1,000 terminated sessions per second for all three tests (in a constant way, constantly making and breaking over the whole remaining test duration of 15 minutes).


Thanks, Carsten

JeddChen 12/5/2012 | 4:24:09 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Emulator is a very important and supplementary means for functionality and performance test. But real networks are quite different and much more complicated and there are a lot of potential and unpredicted factors. That a lot of emulators were used in this test solution has&nbsp; arose the worry mentioned above. so some of the results will be expected to be verified in live networks(trial office).

2skhatri 12/5/2012 | 4:24:04 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Are we to surmise that the upper limit for number of simultaneously attached users is around 1 million?&nbsp;

cross 12/5/2012 | 4:24:03 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Hi tohhwee72,


The primary focus of the ASR1000 Firewall/NAT test was the scale of the stateful network address translations. We did not mean to test the ASR1000 for throughput performance here.&nbsp; Also, the Avalanche emulator is an application-layer device; it sent HTTP requests and received responses on layer 7 - this was not a packet-layer test.&nbsp; To answer your questions:


1. The HTTP requests were for pretty short URLs so the packet sizes were in the order of 150 Bytes.&nbsp;&nbsp; Responses were sent for mid-size objects of a couple of kilobytes; MTU was set to 1470 bytes so there was an average packet size north of 1,000 bytes.&nbsp;


2. Three quarters of the traffic was HTTP, one quarter FTP.


3. The purpose of this test was to validate business security aspects of NAT.&nbsp; Enterprises typically want to hide their internal IP addresses from the Internet.&nbsp; NAT was therefore configured in a 1:1 scenario.&nbsp; 80,000 IP addresses from the 10.0.0.0/8 network were mapped into another 80,000 IP addresses in the (supposedly public) network of 12.0.0.0/8.


4. CPU load was not monitored. The Spirent Avalanche controller reported downstream throughput of 1.09 Gbit/s and upstream throughput of 0.154 Gbit/s. This was not a throughput test - we focused on stateful NAT mapping scalability.


Thanks Carsten

cross 12/5/2012 | 4:24:02 PM
re: Testing Cisco's Mobile Core, Data Center & Business Services

Hi JeddChen,


Network emulations and trials both have their reason for existence I believe.&nbsp; I disagree with your point that trials are better - they just answer different questions at a different time.


First and foremost, the goal of our test was to provide true, real, independent data of a vendor solution publicly.&nbsp; I have not heard of any serious operator trial where results have been made available publicly to the level of detail of this test.&nbsp; I hope our results will be helpful to gauge Cisco's solution for anybody worldwide.&nbsp; It does not take a site visit or good relationships with the trialling operator or vendor to read our article.


Second, it is dangerous to design trials to answer scalability questions. After all, customers are meant to use the trial as a service to some extent. To get to a level where one million users would simultaneously use LTE data connections, one would need to have at least a subscriber base of 10 million hyperactive data users because they are never going to use the network all at once.&nbsp; Few operators will be adventurous enough - or even able, given their customer base - to answer these scalability questions in a trial. TeliaSonera has not disclosed the number of subscribers in its production grade LTE network (not a trial!) - TeliaSonera has just said it has "thousands of subscribers" on its Swedish LTE network, and a total of around 500,000 mobile broadband subscribers nationwide for LTE and 3G together.


On the other hand, our test was unable to answer typical trial questions such as: How do subscribers perceive the service? What happens in the interaction with base stations when a large number of users are on the move, fast or slow, or do not have perfect coverage? How will charging work?&nbsp; So I definitely see the viability of trials once scalability emulations have been completed so it will be safe to deploy a trial network for real customers.


Apologies for the long post.


Thanks Carsten


&nbsp;

Page 1 / 2   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE