This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012
��-VY^HYKPUN�WLYMVYTHUJL�Y\UUPUN�KPYLJ[S`�VU�OHYK^HYL*VUÄN\YH[PVU!���/;�*HWHISL�*VYLZ����.)�9(4 o 4.2 Mpps at 64 bytes. o Line-Rate at 375 bytes� � � V���4WWZ�040?�[YHMÄJ�^OPSL�KVPUN�(*3Z��\97-�HUK�8V:�
��-VY^HYKPUN�WLYMVYTHUJL�Y\UUPUN�PU�H�]PY[\HS�THJOPUL*VUÄN\YH[PVU!��]*7<����.)�9(4 o Eight Tenants – Two Gigabit Uplinks� � � � � � � ���� �.IWZ�6\[IV\UK o Line-Rate IMIX between Two Ports o uRPF/ACLs have no impact on performance
Vyatta Subscription Edition 6.5 R1
Testing Dates: October 10th – 30th 2012Report Updated: November 20th 2012
Table of Contents ................................................................................ 3!Table of Figures ................................................................................... 3!Results Summary ................................................................................. 4!Test Lab Setup .................................................................................... 5!Tests Carried Out ................................................................................. 6!Testing Results Reviewed .................................................................... 7!Appendix A – Test Cases ................................................................... 12!Appendix B – Sample Test Result Data Tables ................................... 14!About Router Analysis ........................................................................ 15!
Table of Figures Figure 1 - Vyatta 6.5 Deployment Scenarios. ..................................... 4 Figure 2 – Vyatta 6.5 R1 VM Setup ..................................................... 5 Figure 3 - Vyatta 6.5 R1 IPv4 and IPv6 IMIX Forwarding .................... 7 Figure 4 - ACL + uRPF Results ........................................................... 8 Figure 5 - QoS Loss On TOS 0 at 57% Line Rate ............................... 9 Figure 6 - QoS Loss on TOS 0 at 57% Line Rate ................................ 9 Figure 7 - At 69% TOS 4 is affected ................................................. 10 Figure 8 - 39% Loss On TOS 4 ......................................................... 10 Figure 9 - 500k BGP Routes ............................................................. 11
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
Results Summary Vyatta Subscription Edition 6.5 R1 (Vyatta SE 6.5) performed better than
expected when tested in the Router Analysis test lab. Tests were run on both Bare Metal (directly on the server) and in a Virtual Machine (under VMWare ESXi 5.1).
Featureless IPv4 forwarding performance for 64 byte packets crossed
4Mpps after modifications were made to the host server used for the test. Initial non-optimal setup tests showed 1.7Mpps.
Enabling features such as ACLs and uRPF dropped the routers forwarding performance for 64 byte packets down to 3Mpps for IPv4.
QoS functioned as expected, protecting streams as configured. Below are the deployment scenarios that were tested by Router Analysis.
Testing included two different VM setups, one using two vCPUs, one using four vCPUs and a Bare Metal setup using the test lab setup. The following graph is the throughput of each system while running IMIX traffic over it.
Figure 1 - Vyatta 6.5 Deployment Scenarios.
!
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
Test Lab Setup The Router Analysis test lab utilizes specially built systems from IXSystems
to host software based systems. The hardware specs are the following: SuperMicro X9SAE-V with a single Intel I7-3770 CPU, 32GB ECC RAM,
240GB Intel 520 Series SSD and 4 Intel I340-T2 GbE NICs. Setting up the system to optimally forward packets is a key part of working
with any software forwarding router. The choice of network cards was key to getting the results included in this test.
Vyatta SE 6.5 was configured to use all eight GbE interfaces. Tests were
done with 4GB RAM on Bare Metal but Vyatta never used more than 1GB of ram during any tests.
When testing under VMWare, Vyatta SE 6.5 was given 4GB RAM, 2vCPUs,
vSwitch access to six interfaces and DirectPath access to two interfaces. A total of eight tenant systems were emulated including two on the same VMWare host and six via links to the tester.
To generate data for the 4vCPU test, four VMs were used along with four
GbE interfaces for tenants. The test used four GbE interfaces connected via DirectPath for uplinks.
Figure 2 – Vyatta 6.5 R1 VM Setup
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
The main takeaway from the results is that Vyatta SE 6.5 performs well both
in a Bare Metal environment and as a Virtualized Multi-Tenant Router. While testing, the lowest performance appeared while executing the “IPv6
forwarding with ACLs, uRPF, and QoS IMIX” test case. Even then the performance was reasonable at 2.25Mpps.
The biggest impact within virtualization was using the standard vSwitch to
connect hardware interfaces. Future tests may include using the distributed vSwitch or specialized hardware to see how much impact the different vSwitch types have. !
0
500000
1000000
1500000
2000000
2500000
3000000
IPv4 Base IPv4 ACL IPv6 Base IPv6 ACL
2vCPU VM
4vCPU VM
Bare Metal
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
Base Test Cases: IPv4/IPv6 IMIX Forwarding + Features Tests For these tests we follow the following testing concepts to find the best-case performance of the system. Testing is done using IMIX traffic on bare-metal. Step 1: Configure 8G of ports and send traffic in a full mesh – 2.58Mpps.
Step 2: Configure a simple permit all ACL on the router – No Change. Step 3: Configure the ACL inbound on one interface – 2.481Mpps. Step 4: Configure the ACL outbound on one interface – No Change.
Step 5: Configure the ACL in/out on all interfaces – 2.38Mpps. Step 6: Configure traffic to run at about 2.38Mpps. Step 7: Configure uRPF on one interface – No Change.
Step 8: Configure uRPF on a second interface – No Change. Step 9: Configure uRPF on the rest of the interfaces – No Change. Result: uRPF has no impact on traffic.
!Figure 4 - ACL + uRPF Results
!!
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
QoS Testing and Results: From our tests we have determined that QoS works as expected on the
system. In a simple test involving four GbE links (a three to one oversubscription), three different TOS bits were set (0, 4 and 7) and traffic was sent balanced (33.3% per TOS) from each GbE towards a single GbE port outbound.
The router was configured so that TOS 0 was protected up to 15% of line
rate, TOS 4 was protected up to 15% of line rate and TOS 7 was protected up to 70% of line rate.
When running the test the following results were seen: < 100% all streams went through without issue > 100% TOS 0 started to be dropped ~ 140% TOS 4 started to be dropped ~ 190% TOS 7 started to be dropped Step 1: Send less than 33% line-rate from each of the three ports
Result: Up to almost 33% line-rate from the 3 ports (< 100% outbound on
other interface) no loss was seen. Above 33% we saw loss for the TOS 0 stream but not the TOS 4 or 7 streams. Step 2: Send 50% line-rate from the TOS 0 port (113% total)
Figure 5 - QoS Loss On TOS 0 at 57% Line Rate
Result: At above 34% we start to see TOS 0 traffic drop as shown in the
magnified section below where we are sending 50% traffic on the TOS 0 link:
Figure 6 - QoS Loss on TOS 0 at 57% Line Rate
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
The following two test cases represent a skeleton format of the set of tests performed in the Router Analysis lab. Test Case One – IP Forwarding Synopsis: When testing systems that have a set PPS or Forwarding Bandwidth that is lower than the combined interfaces available; it is necessary to validate the best-case base-forwarding rate. Once the rate has been determined, it is possible to judge other test cases against the best-case forwarding rate. Goal of Test: To confirm best-case IPv4 or IPv6 forwarding rate using the minimum and maximum packet size accepted by the system. Procedure: Configure the minimum number of links between the tester and the system under test (SUT) to provide either a.) More than the maximum PPS or b.) More than the maximum forwarding bandwidth as claimed by the vendor. For example: Vyatta SE 6.5 performance is based on the hardware it is installed on. To test the system you would start with the same number of interfaces as cores you have in the system. Data to Col lect: Maximum PPS for IPv4, IPv6 and a 50/50 mix of IPv4 and IPv6 Maximum traffic forwarding for IPv4, IPv6 and a 50/50 mix of IPv4 and IPv6 !
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
Test Case Two – QoS Testing Synopsis: !This is a simple QoS test based on information from the IXIA BlackBook. We have modified it slightly to utilize more than 2 ports. Goal of Test: Determine if QoS functions as expected by creating three different TOS valued streams, 0, 4 and 7. On the router configure a traffic policy to prefer 7 over 4 and 4 over 0. Procedure: Using IxNetwork create three streams with TOS levels of 0, 4 and 7 on three interfaces going towards a single interface. On the outbound interface configure a QoS policy to prefer TOS 7 over TOS 4 and TOS 4 over TOS 0. Run traffic and slowly raise it until you see packets drop for 0, then 4 and then 7. Addit ive Tests: Add in background traffic from your earlier test where you saw packet loss due to features. Confirm that the QoS policy is still working as expected. Data to Col lect: Packet loss data for each stream at different traffic levels. Screen Captures of the IxNetwork screen during the tests. !
RA-111812-02 Vyatta Subscription Edition 6.5 R1 Test Plan and Results November 2012 !
Appendix B – Sample Test Result Data Tables The following tables represent a sample of the data collected during testing in the Router Analysis Lab. IPv4 Forwarding Results Data Bare-Metal Eight GbE Interfaces Packet Size (bytes) Aggregate No Drop Rate (pps) Aggregate TX Rate (Gbps)
With a world-class fully independent testing lab, Router Analysis provides unbiased information to its clients.
Router Analysis provides the following services: Network Equipment Testing Consultation and Advisement Test Plan Creation On-Site Support Network Design Consultation, Verification and Installation … And More. Contact Router Analysis by phone at +1 888 SONN COM (766-6266) or by
email at [email protected] for a quote, more information or any questions.