Alpha Princeton

Show Contents...Hide Contents...

Contents

• Hardware/Software
• Participants
• Reason for interest in NetFPGA
• Role in Alpha Program
• Progress
• Progress on Reference Router Testing
• Feature Testing
  • Testing of packets with MTU between 1400 to 2000 bytes

Hardware/Software

• NetFPGA Board #24

Participants

• Lead: Jen Rexford
• Eric Keller (Alpha user + NetFPGA Developer): ekeller@princeton.edu
• Yi Wang (Alpha user: Added 10/23/07): yiwang@cs.princeton.edu

Reason for interest in NetFPGA

• VINI

Role in Alpha Program

• Testing of NIC and reference router features

Progress

• Date that hardware arrived in lab
  • Sent to Princeton in July, 2007
• Date that hardware passed self-test
  • Reported in Oct 4, 2007 email
• Date that hardware passed NIC regression test
  • Reported in Oct 17, 2007 email
• Date that hardware was synthesized:
  • Reported in Oct 17, 2007 email
• Date that NIC performed ping
  • Oct. 22, 2007
• Alpha Test Contribution
  • Eric Keller and John Lockwood defined test Oct 25, 2007
  • Will examine how NIC processes packets below, at, and slightly above MTU of 1500 bytes
  • Code should be integrated into NIC regression suite
  • Due: Nov 12, 2007

Progress on Reference Router Testing

• Date that Reference router installed (Steps 1-14): Nov 28th, 2007
• Date that Hardware and Software Tested (Step 16):
  • Issues - I don't have a machine with 2 network cards in addition to the NetFPGA card. I ordered one from buy.com, but it was damaged in transit, so it'll be another week.
  • I ran extensive tests with a multi computer setup, using cli.c to setup the IP and ARP tables (note:cli seg faulted on me, I needed to edit common/util.c to add "#include <string.h>" to the top to get it to not seg fault).
• Date that Reference router hardware synthesized (Step 17-19):
  • Nov 29, 2007 - didn't run regression, but ran my own tests.
• Date that GUI Tested: Nov 28, 2007
  • Issues - Didn't work. The error suggest that there is a word width issue, which I'm guessing is because I'm on a 64bit machine and libnf2.so was compiled with a 32 bit machine. I couldn't get the make file to regenerate that file, even deleting it. I didn't put much effort into trying to get it to work since cli and regread/regwrite are sufficient for me.
• Other comments as I track them:
  • Click Here

Feature Testing

Testing of packets with MTU between 1400 to 2000 bytes

Owner

Eric Keller

Description

• send 50000 packets with varied length
  
  • between 1400 bytes to 2000 bytes.
    
• Observe how packets are:
  
  • processed
    
  • dropped
    
  • cause errors.
    
• Check counters for
  
  • dropped packets due to length errors
    

$ Issues:

• NetFPGA Alpha hardware was designed to packets of length = 1514 bytes
  
  • MTU=1500 bytes for packet payload (IP)
    
  • 14 bytes of Ethernet header
    
• NetFPGA Beta hardware should support packets of length up to 2000 bytes
  
  • MTU=2000 bytes for packet IP payload
    
  • 14 bytes of Ethernet header
    
• Alpha to Beta release will require
  
  • Same use of (18 kbit = 2 KBytes) Block RAMs
    
  • Change in core-generated FIFOs to support larger packets
    
  • We do not plan to support 8k Jumbo packets because they will consume large amounts of on-chip memory
    

Invocation

See README in download

Output

See README in download

• Print results
  

Regression suite

• Similar to NIC Alpha Tests 2.2: TX/RX Maximum Length packet
  

Known bugs/issues

• Alpha hardware should support MTU=1500 bytes packets
  
• Beta hardware and software should have MTU=2000 bytes
  

Due date

Nov. 12

Download the test from test_ethtonf2_jumbo.tar.gz

---

Dev,,Alpha,,Beta(ro),,Beta-plus(ro)