Intel Corp IXP425/IXP420 Network Processors
Total Page:16
File Type:pdf, Size:1020Kb
TH E TOLLY GROUP No. 203141 November 2003 Intel Corp. Test Intel® IXP425/IXP420 Network Processors Summary Performance Analysis of 802.11 Broadband Routers and Access Points Premise: The market for small Test Highlights office/home office (SOHO) and small/medium business (SMB) is Intel IXP425-based Linksys WRV54G in most cases delivers more inundated with a range of broadband than four times the wired throughput of other broadband routers wireless gateway products at highly tested in unidirectional LAN-to-WAN TCP throughput tests competitive prices. These devices Intel IXP425-based Linksys WRV54G achieves from 41% to 55% are equipped with features that uti- greater throughput than the other devices tested in unidirectional lize the advantages of present-day WLAN-to-WAN test wireless technology and state- of-the-art network processor Intel IXP425-based Linksys WRV54G performs almost 21% capabilities. It is important for better than the Broadcom-based Linksys WRT54G router while consumers to know the performance delivering WEP-encrypted data in WLAN-to-WLAN testing characteristics of these SOHO Intel IXP420-based D-Link DWL 7000AP outperforms other APs routers/access points in order to tested in 128-bit WEP bidirectional WLAN-to-LAN TCP tests better understand these products’ value propositions and to aid in Unidirectional LAN-to-WAN Wired TCP making informed purchasing Throughput of Wireless Broadband Routers decisions. As reported by NetIQ Chariot 100 ntel Corp. commissioned The ITolly Group to benchmark the 94.7 94.4 performance of several commercial- 80 ly available 802.11 wireless gate- ways (wireless broadband routers 60 and access points) that utilize the Intel IXP425 and the Intel IXP420 40 network processors, and compare 36.4 the performance of those devices 20 TCP throughput (Mbps) 22.7 against other generally available 17.8 products based upon rival chipsets. 0 Tolly Group engineers used the Cross-over Linksys Linksys D-Link DI774 Netgear Cable WRV54G WRT54G FWAG114 industry-accepted NetIQ Chariot’s “throughput.scr” script to determine Intel®IXP425 Broadcom ARM9 Broadcom 4702 the aggregate throughput of the Broadband router tested devices under test. Throughput was Source: The Tolly Group, November 2003 Figure 1 © 2003 The Tolly Group Page 1 The Tolly Group Intel Corp. IXP425/IXP420 802.11g Client Unidirectional WLAN-to-WAN TCP Throughput of Wireless Broadband Routers As reported by NetIQ Chariot 25 23.0 23.3 20 16.6 16.5 15 15.7 15.8 14.7 14.9 10 TCP throughput (Mbps) 5 0 No encryption 128-bit WEP 802.11g (Laptop PC client using Linksys WPC54G wireless card) Linksys WRV54G (Intel®IXP425) Linksys WRT54G (Broadcom) D-Link DI774 (ARM9) Netgear FWAG114 (Broadcom 4702) Source: The Tolly Group, November 2003 Figure 2 recorded in two different sce- 802.11g Client with WPA Unidirectional narios: (1) unidirectional WLAN-to-WAN TCP Throughput of WLAN- Wireless Broadband Routers to-WAN traffic, where the As reported by NetIQ Chariot Endpoint 1 (client) in the wire- 20 less network sends data to Endpoint 2 (client), which is 16.9 wired to the WAN port in the 15 15.8 tested device, and (2) bidirec- tional WLAN-to-WLAN traffic, 12.0 where the data flows in both 10 11.4 directions between Endpoint 1 and Endpoint 2 where both are wireless clients. Engineers mea- 5 sured throughput for devices TCP throughput (Mbps) supporting 802.11b, 802.11g and 0 mixed mode (802.11b/g) wireless Linksys Linksys D-Link DI774 Netgear standards. Extensive testing was WRV54G WRT54G FWAG114 conducted between September Intel®IXP425 Broadcom ARM9 Broadcom 4702 Intel®IXP425 2003 and November 2003. Intel®IXP425 Broadband router tested The test scenarios are indicative Source: The Tolly Group, November 2003 Figure 3 © 2003 The Tolly Group Page 2 The Tolly Group Intel Corp. IXP425/IXP420 of those encountered by wire- Bidirectional WLAN-to- less laptop PC users surfing the WLAN (testing 802.11g Intel Corp. Web with a connection to a with no encryption, broadband wireless router con- 802.11g with 128-bit WEP, nected to a wired infrastructure. 802.11b/g with no encryp- IXP425/IXP420 Moreover, test scenarios also tion and 802.11b/g with Network show how wired PCs connected 128-bit WEP) Processors to the router do not hit any bottlenecks when utilizing For the baseline unidirectional TCP Web services. LAN-to-WAN TCP throughput Throughput test, the Linksys WRV54G Test results show that the (with the embedded Intel Intel Corp. wireless gateways based upon IXP425) delivered 94.4 Mbps the Intel IXP425 and IXP420, IXP425 Network Processor of throughput to a wired client Product Specifications* respectively, either outperform over a traditional 100-Mbps competitive products in Fast Ethernet link, more than Member of the IXP4XX network processor most scenarios, or provide five times the throughput of product line for enterprise, small-to-medium performance on par with other the Netgear FWAG114 and 2.5 enterprise (SME), residential and other devices. times the throughput of the networking applications D-Link DI774, the closest com- Intel® XScale® RISC core at up to 533 Results MHz provides headroom for customer- petitive device. (See Figure 1.) defined applications Broadband Router Integrated hardware acceleration of For the unidirectional 802.11g popular cryptography algorithms (SHA-1, Throughput WLAN-to-WAN TCP through- MD5, DES, 3DES, AES) for secure put test, results tracked almost applications On the broadband router side, identically between the baseline IPsec-enabled Network Processor Engine The Tolly Group tested four performance test and the 128-bit (NPE) to accelerate cryptography and products, starting with an Intel WEP test. Here, again, the authentication algorithms IXP425-based Linksys Intel IXP425-based Linksys DSP software library on the Intel XScale WRV54G. Other broadband core supports two to four voice channels WRV54G outperformed all routers tested were: D-Link and reduces system cost other products. In the 128-bit DI774 (using an ARM9 Two high-speed serial (HSS) ports for WEP scenario, the WRV54G VoIP SLIC/CODEC or T1/E1 chipset), a Linksys WRT54G delivered 23.3 Mbps of aggre- Two integrated 10/100 Base-T Ethernet (using an unknown Broadcom gate TCP throughput, which MACs with MII interface for design chipset) and a Netgear represents anywhere from 41% flexibility and cost-effective wire-speed FWAG114 (using a Broadcom performance to 56% greater throughput than BCM 4702 chipset). All four UTOPIA 2 interface with multiple the other devices tested. (See devices were subjected to three ADSL/G.SHDSL or VDSL support Figure 2.) It is also interesting TCP throughput tests: 33/66-MHz PCI v2.2 host and option to note that the similarity interface for glueless connection of up to Unidirectional LAN-to- between the baseline test results four devices WAN (baseline between and the 128-bit WEP results Low system power consumption (1.0 to 1.5 watt typical) two computers connected shows that the WEP encryption via a crossover cable) does not adversely impact For more information contact your Intel Unidirectional WLAN-to- bandwidth performance. sales representative: WAN (testing 802.11g with URL: http://intel.com/design/network/ Engineers also tested the no encryption, 802.11g products/npfamily/ixp425.htm devices in the unidirectional with 128-bit WEP, and 802.11g WLAN-to-WAN test *Vendor-supplied information not verified by 802.11g with WPA and a scenario using the recently The Tolly Group pre-shared key with TKIP) introduced Wi-Fi Protected © 2003 The Tolly Group Page 3 The Tolly Group Intel Corp. IXP425/IXP420 802.11g Client Bidirectional WLAN-to-WLAN TCP Throughput of Broadband Routers As reported by NetIQ Chariot 10 9.7 9.4 9.1 8 8.5 8.0 7.3 6.8 6 6.6 4 TCP throughput (Mbps) 2 0 No encryption 128-bit WEP 802.11g (Laptop PC client with Linksys WPC54G interface) Linksys WRV54G (Intel®IXP425) Linksys WRT54G (Broadcom) D-Link DI 774 (ARM9) Netgear FWAG114 (Broadcom 4702) Source: The Tolly Group, November Figure 4 Access (WPA), a standards- Linksys WRT54G device while 3Com LAN AP8200, a Cisco based security mechanism that delivering WEP-encrypted Aironet 1220 (using an IBM eliminates most 802.11 security data. (See Figure 4.) For non- Power PC 405) and a Broadcom- issues. Maximum throughput encrypted throughput, the score based Linksys WAP54G. All for the Intel IXP425-based difference was about 29% for four devices were subjected to Linksys WRV54G was 16.9 these devices, with throughput a bidirectional 802.11b WLAN- Mbps, or about 7% higher of 9.4 Mbps and 7.3 Mbps to-LAN TCP throughput test. than the D-Link DI774. Both respectively. Even in the mixed Broadcom-based devices exhib- mode test (802.11b/g), the Here, the D-Link DWL 7000AP ited lower throughput values Linksys WRV54G offered any- with the Intel® IXP420 per- with the least value observed where from an 18% to 48% formed on par with the Cisco in the case of the Netgear performance gain over other Aironet 1200 series (1220) FWAG114; the Intel IXP-based devices tested. (See Figure 5.) access point in delivering the router delivered 48% greater 802.11b non-encrypted traffic to throughput. (See Figure 3.) Access Point an 802.11b Centrino client with Throughput rates of 5.1 Mbps and 5.3 Mbps In an 802.11g bidirectional respectively. In the case of 128- WLAN-to-WLAN test, the The Tolly Group tested four bit WEP-encrypted traffic, Intel Linksys WRV54G with the products, starting with an Intel IXP420-based DWL 7000AP Intel® IXP425 performed about IXP420-based D-Link DWL outperformed the others with a 18% better with a rate of 9.7 7000AP access point.