Extreme S450 vs Cisco 3750G Satcking Review

Extreme S450 vs Cisco 3750G Satcking Review

DRAFT COPY REV-01

Summary: Our testing shows that the Summit X450a-48t offers superior Gigabit Ethernet network performance over the Cisco 3750G. The stacked Summit x450a-48t, running ExtremeXOS and SummitStack modular switch OS, can better support faster throughput for end-user applications.

Extreme Networks Summit X450a-48t

o Predictable behavior when throttled to its stated

throughput.

o Higher throughput performance at both layer 2 and

layer 3.

o Superior stack-master redundancy.

Cisco Catalyst 3750G-48TS

o Frame loss and loss of performance when the

master fails in a layer 3 scenario.

o Lower throughput performance between stacks. o Inferior stack failure technique.

Test Equipment: Extreme Networks:

2 Stacked Summit X450a-48t Software version 12.0.0.7 Cisco Networks:

2 Stacked Cisco 3750G-48TS Software version: 12.2.25(SEC) Ixia:

Ixia Optixia

IxOS 4.10

IxAutomate 6.10 IxNetwork 5.10

Test 1: Layer 3 Throughput Performance on “Backup” Switch

Objective: The purpose of this test is to determine the throughput performance on a stacked ‘backup’ switch. The test would also determine the maximum amount of throughtput supported in the Extreme Networks Summit X450a-48t and the Cisco Catalyst 3750G-48TS.

Methodology: Using Ixia’s IxExplorer, the test system creates aggregated throughput traffic to its corresponding ports and VLANs.

The Ixia tester is configured to send unicast traffic using 26-40 GbE ports of the switch. The number of tester ports is determined by the supported limits of the switch. Three tests were created from the main objective.

1. 2 Stacking Cables Connected

2. Removing 1 Stacking Cable (Simulating

Link Failure)

3. Failed Master Scenario (0 Stacking Cables

Connected)

Results:

On the first test, the Cisco 3750G forwarded 13 Gbps full duplex layer 3 traffic without loss at 38.7Mpps of throughput. On the second test, where a stacking cable is removed, the total amount of throughput performance is cut in half to 50%. In the last test where the master would fail, the Cisco 3750G would change its stack MAC address and fail the test.

In the charts below, the theoretical bandwidth of 13 Gps full duplex is 40.5Mpps. During the test of the Cisco 3750G maintained a throughput of 38.7Mpps with two stacking cables connected. With one stacking cable, the bandwidth of the stack was cut in half to 18.2Mpps. On the last test, the switch cannot pass traffic due to the stack MAC address changing with the election of a new master.

The alarming behavior in the Cisco’s 3750G method of stacking is the layer 3 centralized routing at the Master node. If the stack master were to fail, the election process will start (for about 1-2 seconds), then it would issue a new stack MAC address. In result, the layer 3 routing would be centralized at the master, and could compromise applications and users who do not accept gratuitous ARP’s.

When performing the same tests, but with 40 Gbps, the Summit X450a-48t performed as expected.

On the first test the Summit X450a forwarded 20 Gbps full duplex layer 3 traffic without loss at 59.5Mpps of throughput. On the second test, the Summit X450a performed exceptionably well without any frame loss or loss in throughput. For the master failure test, the Summit X450a kept all the traffic running at 100% line rate without any loss to the throughput. In the chart below, the Summit X450a maintained a steady amount of throughput during the duration of the test.

Extreme S450 vs Cisco 3750G Satcking Review相关文档

最新文档

返回顶部