Nokia produced a Tolly report it sponsored that compared Juniper’s NetScreen-5400 against its own IP2250. The report says Nokia’s box has three times the HTTP throughput, twice the firewall throughput, and five times the connection rate.
CEO Kevin Tolly said in a statement: Considering that there is not a huge chasm in the pricing spread of the devices tested, users will be able to wring greater performance, and ultimately more bang for the buck from the Nokia IP2250.
The report found that the IP2250, a Nokia device running a firewall from Check Point Software Technologies Ltd had an average throughput of 1.1Gbps on mixed-size HTTP objects, compared to the 5400’s average of 328Mbps.
The report also found that the Nokia box handled 7.7Gbps of firewall traffic, compared to Juniper’s 4.1Gbps, when hit with a torrent of 1,518-byte Gigabit Ethernet frames. That ratio was similar on 512-byte frames and Nokia was 10% faster on 64-byte frames.
A third test showed that the Nokia could handle 32,900 TCP sessions per second, compared to the Juniper box’s 6,920.
Juniper director of product management Lee Klarich said that some of the results are misleading, some did not match what Juniper found when it attempted to replicate the results, and that some of the NetScreen hardware tested was outdated.
They tested one of a possible three processing blades against a fully loaded Nokia, Klarich said. The NetScreen 5400 chassis has room for three processing blades, but the Tolly test used just one, he said.
The Tolly test also used an older, slower management blade, Klarich said. The tests were conducted starting in August, three months after a newer faster blade was available, he said.
Klarich admitted, however, that a fully loaded 5400 would make the device much more expensive that the IP2250. All-inclusive, a three-blade would run for about $200,000, compared to the $130,000 configuration it appears Tolly used.
So Tolly’s assertion that the test shows the IP2250 offers more bang for the buck seems accurate. But Klarich also claimed that the tests could be easily tweaked, without changing the hardware, to provide the opposite results.
Usually these test are designed to make their [the sponsor’s] products look good, or they use wrong configurations or wrong products. I know this because we used to, many years ago, Klarich said. Juniper no longer uses Tolly, Klarich added.
