John,
Thank you for the information. I am now much less confused. It seems
what I'm looking for is not 802.1q VLAN tagging, so much as
802.1ad/802.3ad link aggregations as in most of my BigIP configurations I
only have 1 VLAN per interface.
I understand what you're talking about regarding sessions using the same
link in the trunk. From what I've read of Cisco's implementation of
802.3ad (Fast/Gigabit Etherchannel) the switch determines what port in the
trunk to switch the packet to by performing an XOR of the source and
destination MAC addresses.
A question: If I'm becoming concerned about bandwidth in and out of my
BigIP's, would you think I'd be better off trunking 4 100Meg ports, or
replacing my front and/or back interfaces with gig interfaces? Last I
saw, I believe the gig cards sold by F5 can do somethingl like 340Megs
tops, so from a strictly mathematical standpoint I'd get a bit more out of
a 400meg trunk, but I'm wondering if there are any other factors I should
be thinking about.
Thanks again.
-Tim
On Wed, 16 Jan 2002, John Hall wrote:
>
> Trunking is unfortunately an over-used word, causing much confusion
> in the world today... :(
>
> Your reference to 802.1q leads me to believe you are talking about the
> trunks that are defined 802.1q Annex D where you have multiple tagged
> VLAN's configured on a single interface in order to pass traffic for
> multiple VLAN's between two switches. All BIG-IP versions after v3.1
> support tagged VLAN's (although I have to admit the support got *MUCH*
> better after v4.x) and assigning multiple VLAN's to a single interface
> for the purpose of VLAN trunking (that's what *I* call this use).
> Although I'm confused by your reference to "bandwidth aggregation".
>
> Rick has already replied to you referring to our "bigpipe trunk"
> command, which is used to configure 802.1ad Link Aggregations, which
> are used to make multiple links (physical ports) operate as a single
> interface providing the sum of the bandwidths of the aggregated links
> to the interface and also providing link failure protection (if one
> of several aggregated links goes down, the others continue to operate).
> One practical note is that, like most 802.1ad implementations, ours
> works most efficiently if the number of interfaces configured is a
> multiple of two. Another quirk of 802.1ad that is often not known
> is that all packets associated with a single IP connection (say a
> single telnet session) are *always* sent on the same link of an
> aggregation (this is required by the spec). So, if you setup a
> link aggregation, using 4 100BaseTX ports, and try to download a
> single huge file via FTP, that single download will only be allowed
> approximately 100Mb/s of bandwidth, but if you have a bunch of sessions
> going simultaneously, you will see all four links utilized.
>
> Before you ask, yes, you can configure a single set of links (say four
> 100BaseTX ports) to operate as an 802.1ad link aggregation (giving you
> close to 400Mb/s of bandwidth on the interface) with multiple tagged
> VLAN's going across that interface to another switch. One of our test
> configurations uses four gigabit interfaces aggregated to a single
> interface over which we pass traffic from four (two internal and two
> external) VLAN's and it works like a charm. Link aggregation allows
> you to use the full bandwidth capacity of your BIG-IP in a balanced
> way when your traffic load is asymmetric (such as with web traffic
> where the incoming requests are small, but the responses are large).
>
> Hope this helps,
> JMH
>
>
> Tim Maestas wrote:
> >
> > I was wondering if anyone has had any experiences (good, bad, or
> > otherwise) with using trunking under BigIP 4.x, in particular into a Cisco
> > 6500 series switch. In reading the 4.x documentation, there is no mention
> > of 802.1q trunk modes like there was in 3.x, so I'm not sure how I would
> > be able to make my Cisco treat my trunked interfaces as a group. In 4.x
> > it seems like outgoing packets from the BigIP would be
> > "load-balanced" between the interfaces, but I don't see how this is
> > possible for incoming packets, without something like 802.1q. Granted,
> > if nothing else trunks will give me interface failover, but I'm looking
> > more for bandwidth aggregation. Any info would be appreciated.
> >
> > -Tim
> >
> > ____________________
> > The Load Balancing Mailing List
> > Unsubscribe: mailto:majordomoIZZATvegan.net?body=unsubscribe%20lb-l
> > Archive: http://vegan.net/lb/archive
> > LBDigest: http://lbdigest.com
> > MRTG with SLB: http://vegan.net/MRTG
> > Hosted by: http://www.tokkisystems.com
>
____________________
The Load Balancing Mailing List
Unsubscribe: mailto:majordomoIZZATvegan.net?body=unsubscribe%20lb-l
Archive: http://vegan.net/lb/archive
LBDigest: http://lbdigest.com
MRTG with SLB: http://vegan.net/MRTG
Hosted by: http://www.tokkisystems.com
This archive was generated by hypermail 2b30 : Wed Jan 16 2002 - 23:10:51 EST