Slow TCP performance for traffic routed by ACE module

Question

Hi,
the customer uses two ACE20 modules in active-standby mode. The ACE load-balances servers correctly. But there is a problem with communication between servers in the different ACE contexts. When the customer uses FTP from one server in one context to the other server in other context the throughput through ACE is about 23 Mbps. It is routed traffic in ACE:-(  See:
server1: / #ftp server2
Connected to server2.cent.priv.
220 server2.cent.priv FTP server (Version 4.2 Wed Apr 2 15:38:27 CDT 2008) ready.
Name (server2:root):
331 Password required for root.
Password:
230 User root logged in.
ftp> bin
200 Type set to I.
ftp> put "|dd if=/dev/zero bs=32k count=5000 " /dev/null
200 PORT command successful.
150 Opening data connection for /dev/null.
5000+0 records in.
5000+0 records out.
226 Transfer complete.
163840000 bytes sent in 6.612 seconds (2.42e+04 Kbytes/s)
local: |dd if=/dev/zero bs=32k count=5000  remote: /dev/null
ftp>
The output from show resource usage doesn't show any drops:
conc-connections              0          0     800000    1600000          0
  mgmt-connections             10         54      10000      20000          0
  proxy-connections             0          0     104858     209716          0
  xlates                        0          0     104858     209716          0
  bandwidth                     0      46228   50000000  225000000          0
    throughput                  0       1155   50000000  100000000          0
    mgmt-traffic rate           0      45073          0  125000000          0
  connections rate              0          9     100000     200000          0
  ssl-connections rate          0          0        500       1000          0
  mac-miss rate                 0          0        200        400          0
  inspect-conn rate             0          0        600       1200          0
  acl-memory                 7064       7064    7082352   14168883          0
  sticky                        6          6     419430          0          0
  regexp                       47         47     104858     209715          0
  syslog buffer            794624     794624     418816     431104          0
  syslog rate                   0         31      10000      20000          0
There is parameter map configured with rebalance persistant for cookie insertion in the context.
Do you know how can I increase performance for TCP traffic which is not load-balanced, but routed by ACE? Thank you very much.
Roman

Answer 1

Default inactivity timeouts used by ACE are
icmp 2sec
tcp 3600sec
udp 120sec
With your config you will change inactivity for every protocol to 7500sec.If you want to change TCP timeout to 7500sec and keep the
other inactivity timeouts as they are now use following
parameter-map type connection GLOBAL-TCP
set timeout inactivity 600
parameter-map type connection GLOBAL-UDP
set timeout inactivity 120
parameter-map type connection GLOBAL-ICMP
set timeout inactivity 2
class-map match-all ALL-TCP
match port tcp any
class-map match-all ALL-UDP
match port tcp any
class-map match-all ALL-ICMP
match port tcp any
policy-map multi-match TIMEOUTS
class ALL-TCP
connection advanced GLOBAL-TCP
class ALL-UDP
connection advanced GLOBAL-UDP
class ALL-TCP
connection advanced GLOBAL-ICMP
and apply service-policy TIMEOUTS globally
Syed Iftekhar Ahmed