[NANOG] Introducing latency for testing?

So I want to mimic some latency in a test network for DB replication.
I am wondering what other's have used for this? Obviously, the best
way to would be to actually have one box across the US or across the
globe to actually test against but what if you don't have that? Are
there any GPL software router solutions that would allow you to tweak
the latency in between the two test boxes?

Thanks in advance.

-Mike

setup a linux box between the systems with netem (included in most distros). http://www.linux-foundation.org/en/Net:Netem
with it, you can introduce latency, loss, jitter, etc.
tate

Mike Lyon wrote:

NISTnet at http://snad.ncsl.nist.gov/nistnet/

Also, there is a commercial (reasonably priced) product called network
nightmare. (http://networknightmare.net/)

Cisco also has an .iso that they'll give to customers that's a NISTnet
livecd.

-Geoff

a message of 15 lines which said:

So I want to mimic some latency in a test network for DB replication.
I am wondering what other's have used for this? Obviously, the best
way to would be to actually have one box across the US or across the
globe to actually test against but what if you don't have that? Are
there any GPL software router solutions that would allow you to tweak
the latency in between the two test boxes?

I use and like FreeBSD's dummynet:

http://info.iet.unipi.it/~luigi/ip_dummynet/

Highly recommended.

Or a FreeBSD box with DUMMYNET (runs through IPFW). You can do all of that stuff.

C. Tate Baumrucker wrote:

Thank you all for the wonderful responses!

Cheers,
Mike

The freebsd dummynet driver is all about latency simulation...

http://www.scalabledesign.com/articles/dummynet.html

linux has a netem which can do the same thing

http://www.linux-foundation.org/en/Net:Netem

joelja

Mike Lyon wrote:

Joel Jaeggli (joelja) writes:

The freebsd dummynet driver is all about latency simulation...

http://www.scalabledesign.com/articles/dummynet.html

linux has a netem which can do the same thing

http://www.linux-foundation.org/en/Net:Netem

  dummynet is significantly easier to set up, especially for
  doing things like random packet reordering / packet loss (using the
  'prob' rule of ipfw + delay property with pipes).

  ipfw pipe 42 config bw 1024Kbit/s delay 6ms
  ipfw pipe 666 config bw 64Kbit/s delay 350ms

  ipfw add 10 prob 0.05 deny ip from 1.2.3.4 to any
  ipfw add 10 prob 0.8 pipe 666 ip from A to B
  ipfw add 10 prob 0.5 pipe 42 ip from A to B

  ... and it runs.

It's not free, but at a recent trade show I did see what appeared to be an
affordable unit from Apposite Technologies (apposite-tech.com). And there's
always PacketStorm.

Frank

Frank Bulk - iNAME wrote:

It's not free, but at a recent trade show I did see what appeared to be an
affordable unit from Apposite Technologies (apposite-tech.com). And there's
always PacketStorm.

Frank

From: Mike Lyon [mailto:mike.lyon@gmail.com] Sent: Friday, May 02, 2008 3:13 PM
To: NANOG
Subject: [NANOG] Introducing latency for testing?

So I want to mimic some latency in a test network for DB replication.
I am wondering what other's have used for this? Obviously, the best
way to would be to actually have one box across the US or across the
globe to actually test against but what if you don't have that? Are
there any GPL software router solutions that would allow you to tweak
the latency in between the two test boxes?

Thanks in advance.

-Mike

_______________________________________________
NANOG mailing list
NANOG@nanog.org
http://mailman.nanog.org/mailman/listinfo/nanog

IIRC ipfw can do this using dummynet and the delay directive.

Regards,

  Chris

Chris Marlatt wrote:

Frank Bulk - iNAME wrote:

It's not free, but at a recent trade show I did see what appeared to be an
affordable unit from Apposite Technologies (apposite-tech.com). And there's
always PacketStorm.

Frank

From: Mike Lyon [mailto:mike.lyon@gmail.com] Sent: Friday, May 02, 2008 3:13 PM
To: NANOG
Subject: [NANOG] Introducing latency for testing?

So I want to mimic some latency in a test network for DB replication.
I am wondering what other's have used for this? Obviously, the best
way to would be to actually have one box across the US or across the
globe to actually test against but what if you don't have that?

>

boxes across the globe have the property of being somewhat less deterministic than you'd like if you need repeatability.

IIRC ipfw can do this using dummynet and the delay directive.

it will also do jitter and drop rate...

to wit, it's exactly what is need here.

there are analogous tools for iptables based platforms.

There is one interesting box that you can use
to simulate the delay at 10GE and do a few other things.

http://projects.gtrc.aist.go.jp/gnet/gnet10p3e.html

  - Jared