11-02-2009 06:55 AM - edited 02-21-2020 03:46 AM
I have two sites and connected with MPLS links of 2MB with Cisco 2821 routers. Now a requirement came that two sites will have additional server ( one each location - for data replication purpose) and 75% of the bandwidth needs to be allocated to data replication servers on each site and rest of 25% bandwidth will be utilize for both sides normal traffic.
Is there any additional modules to be added on each router to isolate the traffic. Please let me know.
Thanks,
11-02-2009 09:08 AM
No additional hardware is needed. You will need to configure QoS, here's an example that gives a host 20%.
To limit a single host the configuration can be a little more straight forward. If
10.16.32.120 is the host that you need to limit to 20% of the total
interface bandwidth, and during congestion (when all traffic reaches the
available bandwidth), the configuration should look as follows.
-------------------
access-list 101 permit ip 10.16.32.120 any
!
class-map Altiris
match access-group 101
!
policy-map RestrictAltiris
class Altiris
bandwidth percent 20
!
interface Gi1/1
service-policy output RestrictAltiris
In this scenario if there is excess bandwidth not in use, host 10.16.32.120
can make use of it, but when the link gets congested the router will limit
the traffic for this host to 20% of the total interface bandwidth, and all
the remaining traffic to 80%.
Hope it helps.
11-02-2009 12:01 PM
To expand on the information that Collin provided . . .
If you really want to cap the bandwidth to a class of traffic, i.e. truly limit replication to 75%, you can add a policer or shaper to a class, similar to what Collin shows.
e.g.
policy-map RestrictAltiris
class Altiris
bandwidth percent 20
shape average 1500000
However, like Collin, since CBWFQ guarantees bandwidth allocations, you rarely need to cap bandwidth if you, for instance, insure your other traffic gets the other 25% (i.e. 100% less 75% for replication).
e.g.
policy-map RestrictAltiris
class Altiris
bandwidth percent 75
(NB: BTW, there are other rules pertaining to bandwidth reservations, but prior examples, both Collin's and mine, should suffice.)
Also BTW, since you mention MPLS, and since MPLS often allows multisite communication, if there are more than just these two sites that can communicate with these two sites across their 2 Mbps links, other considerations apply for dealing with such a situation.
PS:
Although software based QoS should meet your stated requirements, there are optional modules to provide "WAN optimization", see http://www.cisco.com/en/US/prod/collateral/contnetw/ps5680/ps6870/product_data_sheet0900aecd8058218c.html.
PPS:
BTW, when Collin describes:
". . . but when the link gets congested the router will limit the traffic for this host to 20% of the total interface bandwidth, and all the remaining traffic to 80%.", this might be just a bit misleading since when the link is congested individual classes aren't really limited they are instead guaranteed their bandwidths as a minimum, not precisely the same thing. For example with 20% and 80% guarantees, either class could still obtain unused bandwidth from the other class while the link is congested.
[edit]
This being true, for something like what you describe, I would recommend using the least amount of bandwidth guarantee that data replication needs. Assuming your other traffic, on average, doesn't consume more than 25% of the bandwidth, you might find even providing replication only 1% for a bandwidth guarantee works fine. Why you would want to do this, by lowering the bandwidth guarantee for replication, you allow other traffic to burst. Such bursting capability normally improves any kind of transactional or conversational applications and doesn't (usually) unduly delay replication.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide