This is the mail archive of the ecos-discuss@sourceware.org mailing list for the eCos project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]
Other format:	[Raw text]

Re: eCos Performance/ eth0,eth1 Routing problem (number of hops!)

From: "Roy E Richardson" <eng_at_play at cox dot net>
To: <mkhoyila at uci dot edu>,<ecos-discuss at sources dot redhat dot com>
Cc: "Matt Jerdonek" <maj1224 at yahoo dot com>
Date: Tue, 13 Sep 2005 12:04:33 -0700
Subject: Re: [ECOS] eCos Performance/ eth0,eth1 Routing problem (number of hops!)
References: <20050913042433.48669.qmail@web33507.mail.mud.yahoo.com>

There are some separate eCos configuration items that may be worth tinkering with. I am in a hurry, and don't have ready access to the eCos config tool, so the exact names of the config items are not included, but the description should be enough to locate each one.

1) the tick timer list organization - sorted by time vs. non-sorted. The default config timing option is to use an unsorted list of timers, with an explanation to the effect that avoiding the overhead of a sorted by expiration time is more efficient. However the unsorted method requires that the entire timing list be scanned for timers to execute each tick. Use of the sorted timing list is recommended.

2) compiler optimization level - our development was also on a powerPC platform, and we chose level 2 for best performance (difference between 2 & 3 is an arbitrary one, but either is much better than 0 or 1).

3) there is no mention of the level of eCos run-time checks that are effective on the test load - if you have not already minimized them ...

PS. I missed the discussion threads that referenced what the caching resolution entailed - the MPC's caching is a confusing, but extremely effective one - if the system startup does not currently fully utilize what's available to the hardware (both read and write), or portions of memory excluded due to occaisional flakiness, then that's another area to investigate - this is from one who has "been there and done that".

----- Original Message ----- From: "Matt Jerdonek" <maj1224@yahoo.com> To: <mkhoyila@uci.edu>; <ecos-discuss@sources.redhat.com> Sent: Monday, September 12, 2005 9:24 PM Subject: Re: [ECOS] eCos Performance/ eth0,eth1 Routing problem (number of hops!)

I don't see any problem with the routing tables.  It's
curious (but not necessarily problematic) that your
ethx interfaces are also the default gateways.

I have a similar setup, except I have ppp instead of
eth1.  I am able to send a ping request to the PPP
interface, and have the ping response exit the
ethernet interface.  So, I'm reasonable confident that
your setup will work.  You can verify that by pinging
the eth1 interface through the eth0 interface.

Make sure that your ping source has a default gateway
of the eth0 and the ping destination has a default
gateway of eth1.  If that's correctly setup, then I
think you're just going to have to dig into the
source.

-- Matt

--- mkhoyila@uci.edu wrote:

Indeed, it was cache enabling problem.

Here is the new numbers which are getting closer but
still needs some work:

Linux: max 26,000 packets (60 bytes) per second
(single burst) with no drops

eCos: max 23,500 packets (60 bytes) per second
(single burst) with no drops

I DO HAVE A SETUP ISSUE which I need help. I want to
route packet coming
in from IP address: 192.168.1.80 to eth0 (IP ROUTE)
to eth1 and eth1 to
route to its destination ip: 192.168.2.80. As you
see the purpose is to
test performance of my driver and eCos vs Linux by
SmartBits. Here is an
snap-shot of my eCos route table:

UCI DEFAULT GATEWAY TABLE
Interface    Gateway IP addr         MAC addr
--------------------------------------------------
--------------------------------------------------

Routing tables
Destination     Gateway         Mask
Flags    Interface
0.0.0.0         192.168.1.8     0.0.0.0         UG
    eth0
127.0.0.0       127.0.0.1       255.0.0.0       UG
    lo0
127.0.0.1       127.0.0.1                       U
    lo0
192.168.1.0     192.168.1.0     255.255.255.0   U
    eth0
192.168.2.0     192.168.2.0     255.255.255.0   U
    eth1

Arp table
IP addr             MAC addr             Flags
----------------------------------------------
192.168.1.8         02:10:12:01:00:08    UHCL
192.168.1.80        00:00:00:00:00:01    UHC
192.168.2.8         02:10:12:01:00:09    UHCL
192.168.2.80        00:00:00:00:00:02    UHC

Interface statistics
IFP: eth0
IP: 192.168.1.8, Broadcast: 192.168.1.255, Netmask:
255.255.255.0
        UP BROADCAST RUNNING MULTICAST MTU: 1500,
Metric: 0
        Rx - Packets: 0, Bytes: 0, Tx - Packets: 0,
Bytes: 0
IFP: eth1
IP: 192.168.2.8, Broadcast: 192.168.2.255, Netmask:
255.255.255.0
        UP BROADCAST RUNNING MULTICAST MTU: 1500,
Metric: 0
        Rx - Packets: 0, Bytes: 0, Tx - Packets: 0,
Bytes: 0
IFP: lo0
IP: 127.0.0.1, Broadcast: 127.0.0.1, Netmask:
255.0.0.0
        UP LOOPBACK RUNNING MULTICAST MTU: 16384,
Metric: 0
        Rx - Packets: 0, Bytes: 0, Tx - Packets: 0,
Bytes: 0


The ip routing between eth0 and eth1 is NOT
happening. I am starting to
think the number of "hops" is the issue. Here is
initialization prints. Is
there any other issue you see. Thanks.

BOOTP[eth0] op: REPLY
       htype: Ethernet
        hlen: 6
        hops: 0
         xid: 0x0
        secs: 0
       flags: 0x0
       hw_addr: 02:10:12:01:00:08
     client IP: 192.168.1.8
         my IP: 192.168.1.8
     server IP: 192.168.1.8
    gateway IP: 192.168.1.8
  options:
        subnet mask: 255.255.255.0
       IP broadcast: 192.168.1.255
            gateway: 192.168.1.8
BOOTP[eth1] op: REPLY
       htype: Ethernet
        hlen: 6
        hops: 0
         xid: 0x0
        secs: 0
       flags: 0x0
       hw_addr: 02:10:12:01:00:09
     client IP: 192.168.2.8
         my IP: 192.168.2.8
     server IP: 192.168.2.8
    gateway IP: 192.168.2.8
  options:
        subnet mask: 255.255.255.0
       IP broadcast: 192.168.2.255
            gateway: 192.168.2.8


--
Before posting, please read the FAQ:
http://ecos.sourceware.org/fom/ecos
and search the list archive:
http://ecos.sourceware.org/ml/ecos-discuss

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss


--
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss

References:
- Re: eCos Performance/ eth0,eth1 Routing problem (number of hops!)
  - From: Matt Jerdonek

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]