This is actually a request/question for netapp, but I'm hoping to generate a few "me too"'s from the list.
We're a large (1200 systems) ATM shop. For various reasons, we're using Fore SPANS signalling, and will be for awhile. The move to standards-based ATM is going to be ugly, and we (and the standards) aren't quite ready yet.
NetApp supports SPANS ATM, using Fore's PCI card and a port of the SPANS protocol engine in their ATM drivers. This is good.
But, there are two features of Fore SPANS that NetApp doesn't support, that we rely on. This message is a request to NetApp to consider adding support for these two features:
1. load balancing. Fore SPANS on other platforms (IRIX, Solaris) provides interface aggregation such that multiple ATM interfaces on the same host are used equally. This is connection-based load balancing, in that new connections are assigned round-robin to all the interfaces in a given load balance group, on a given machine.
On some platforms, all interfaces in a given system can have the same IP address, which simplifies management, and means there is one "view" of a multi-interface machine from the outside. Routing gets simpler, too.
We can get some balancing using static routing configuations that spreads the load over all the interfaces in a system, but it's not quite the same. Plus, support for load balancing may help provide support for:
2. fail over. If an interface, or switch fabric, or other component fails, Fore SPANS will fail an interface, and all the connections will be re-established on the other available interfaces. Our core switch architecture is such that each server has multiple ATM "feet", in different switches and backplanes, so we've got fault tolerance with automatic failover when somthing in the network path croaks.
It's the failover that we're most interested in, for the fault tolerance and redundancy reasons. load balancing makes administration easier, and helps us better utilize the switch architecture we have in place.
So, NetApp, what are the chances of getting SPANS fail over and load balancing in your ATM implementation?
-Skottie
+--- In a previous state of mind, Scott Miller skottie@fa.disney.com wrote: | | On some platforms, all interfaces in a given system can have the same IP | address, which simplifies management, and means there is one "view" | of a multi-interface machine from the outside. Routing gets simpler, too.
Sounds like etherchannel... I think Alteon was trying this with their Gigabit stuff; unfortunately, no other vendor supports this feature.
| It's the failover that we're most interested in, for the fault tolerance | and redundancy reasons. load balancing makes administration easier, and | helps us better utilize the switch architecture we have in place. | | So, NetApp, what are the chances of getting SPANS fail over and load balancing | in your ATM implementation?
I had always wondered if anybody was using (packet-slicing) ATM for their filers. What sort of performance are you getting?
Alexei
Actually, etherchannel is supported by Sun, Auspex, and Cisco. There may be others, but those are the one that I know of.
David Ritch
Alexei Rodriguez wrote: [...]
Sounds like etherchannel... I think Alteon was trying this with their Gigabit stuff; unfortunately, no other vendor supports this feature.