Duplicate SSH session in Linux

Add 2 lines to /etc/ssh/ssh_config or to ~/.ssh/config:
ControlMaster auto
ControlPath ~/.ssh/socket-%r@%h:%p

You have to login for the first time, the first session will initiate a Control Master by creating a file at ~/.ssh indicate the login name, host and port. After that you can duplicate a ssh session by just execute a ssh command the same way you did for the first login, but this time password is not required.

Original is taken from here:

Advertisements

Huawei made his own GNS3!

labnario

Huawei, to meet your expectations, recently announced the launch of an enterprise network simulation platformeNSP for ICT practitioners. This information you can find at official Huawei website.

To be honest, I have not tested it yet. If I find time I will do it this week. I have already installed this simulator on my notebook and it looks promising. We can build a network based on AR1200 routers and enterprise switches.

View original post 63 more words

S5300HI interface buffers. Buffering issues.

Here is the summary information that I got from Huawei GTAC.
The switch has no any ingress buffers. The whole switch has 2.5 MB of interface buffer space for egress buffering. Each 10GE interface as well as 1GE has 8 egress queues. By default each queue has 6.25KB of buffer statically alocated to them, so summary we have 50KB per each interface.
If you install an extension card into S5300 then new interfaces will share the same common buffer pool and grab 50KB per each interface. Even when interface is shut down it still holds it’s 50 KB reserved.

Since the size of default interface buffer is quite small there can be buffering issues leading to packet drops when the buffer space is exhausted. You can release buffer space that is hold by unused interfaces:

interface GigabitEthernet0/0/5
 shutdown
 qos queue 0 length 0
 qos queue 1 length 0
 qos queue 2 length 0
 qos queue 3 length 0
 qos queue 4 length 0
 qos queue 5 length 0
 qos queue 6 length 0
 qos queue 7 length 0

This configuration returns 50KB of buffer space into common pool. Then you can increase queue length on a highly congested or bursty interface:

interface GigabitEthernet0/0/2
 qos queue 0 length 500000

It is possible to know the size of buffer space available in common pool:

sys
Enter system view, return user view with Ctrl+Z.
[hua-hua]int g0/0/10
[hua-hua-GigabitEthernet0/0/10]qos queue 0 length 1000000000 
Error: This queue support the length range of (0~1134208)
[hua-hua-GigabitEthernet0/0/10]

I intentionally try to configure huge buffer size and get CLI hint revealing real buffer space available: 1134208.

D-ITG seems to be much powerful and efficient than iperf

Official website: http://www.grid.unina.it/software/ITG/
It can provide much more bandwidth without consuming much CPU. Here is simple example.
Receiver:

./ITGRecv

Sender:

ITGSend -a  -sp 9400 -rp 9500 -C 100 -c 1000 -t 30000 -x log.txt 

This will use UDP source port 9400, dst port 9500, send 100pps each 1000bytes long during 30000ms. After testing it will create log file log.txt on the receiver side.
Analyze logs:

./ITGDec log.txt

Useful Huawei OIDs

.1.3.6.1.2.1.47.1.1.1.1.7 – provides names for hwEntityId
1.3.6.1.4.1.2011.5.25.31.1.1.1.1.11 – temperature sensors
.1.3.6.1.4.1.2011.5.25.31.1.1.1.1.12 – temperature sensors high threshold
.1.3.6.1.4.1.2011.5.25.31.1.1.1.1.16 – temperature sensors low threshold
.1.3.6.1.4.1.2011.5.25.31.1.1.1.1.22 – transceiver optical power rx (NE40E)
.1.3.6.1.4.1.2011.5.25.31.1.1.3.1.8 – transceiver optical power rx (switches)