Atlanta - IBM and Cisco Systems, Inc. are moving their battle for the corporate backbone off the street and into the test lab.
At last week's Share user group meeting, IBM passed out performance test results comparing its 2216 mainframe channel-attached Multiaccess Connector and Cisco's channel-attached router, the 7500-based Channel Interface Processor (CIP). Both boxes connect LAN, SNA and TCP/IP traffic to the mainframe via a mainframe channel connection.
The test, audited by The Tolly Group, an independent testing facility in Manasquan, N.J., showed that IBM's 2216 significantly outperformed the CIP in SNA and TCP/IP mainframe channel throughput.
Cisco executives dismissed the test, calling it more of a test of adapter throughput than platform performance.
"We challenge IBM to do this kind of testing totally out in the open in a public forum," said Jeff Graham, CIP product manager for Cisco. "We were not consulted on this test, and it's not a very open way for users to get any meaningful results."
For customers, the likely result of this latest round of sparring will be increased confusion as they try to determine the best way to migrate legacy equipment to new multiprotocol backbones.
The test bed at IBM's Washington Systems Center, in Gaithersburg, Md., included a 2216 Model 400 supporting dual Enterprise Systems Connection (ESCON) -based mainframe channels and a token- ring LAN. The Cisco configuration used a model 7507 router outfitted with a CIP-2 processor supporting dual ESCON attachments and a token-ring LAN, according to IBM.
Jim Goethals, an enterprise solutions manager with IBM, said the test bed supported about 48 servers and a number of traffic generators to build real-life traffic simulations, such as 1k-to-16k TCP/IP file transfers and standard SNA batch transfers.
Tests were first conducted by IBM then by The Tolly Group.
The tests fly in the face of recent Cisco tests. Those compared a CIP with IBM's 3745/3746 front-end processors (FEP) - the box Cisco competes with most often.
Those tests, which were not independently audited, showed that the CIP2 can send traffic to and from the mainframe at 5,500 frame/sec, compared to 3,600 frame/sec for the 3745-31A and 1,200 frame/sec for the 3746-900. Using a 170- byte transaction, the CIP2 handled 6,100 frame/sec, while the 3745-31A handled 3,600 frame/sec and the 3746-900 1,200 frame/sec.
The tests also showed the CIP2 could support nearly four times the amount of SNA sessions as the FEPs. IBM dismissed that test as being "too LAN-centric and not indicative of typical SNA configurations."
Apply for your free subscription to Network World. Click here. Or get Network World delivered in PDF each week.
![]()
Request a reprint or permission to use this article.
