202 ms·

But, can you toss the connections from both providers into a switch and make both avaialable for use all the time? Like a Active-active setup?? Any reason not

But, can you toss the connections from both providers into a switch and make both avaialable for use all the time? Like a Active-active setup?? Any reason not to do that?

mansilladev 8y ago
And put my remote hands worker out of a job? No way.
- caust1c 8y ago
  Yeah, but they must cost you a bundle!
  - mansilladev 8y ago
    That, my friend, is a sunk cost.
    - paul_milovanov 8y ago
      See: sunk cost fallacy
    - cardiffspaceman 8y ago
      So you are committed to paying for the remote hands and can't get out of it, or have already fully paid for lifetime support from the remote hands and can't get a refund :) That doesn't mean that there can't be cost savings from automation. For example if it costs X in lost business due to a misunderstanding by the remote hands that extends the outage unnecessarily, then a certain number of times avoiding that X cost would pay for the investment in automation. You pay the "sunk" and the new cost but you avoid unnecessary costs in the long run. It's all a matter of fully modeling your costs and benefits. Noting that certain costs are sunk is a partial model.
TimTheTinker 8y ago
From 5 minutes of Wikipedia reading, it appears this requires special support from the router to enable “sticky sessions” which prevent out-of-order packets; or, the device OS itself can “stripe” packets across NICs with special (extra) software to enable that.
- XorNot 8y ago
  ECMP handles this just fine by default with Linux (it's per flow aka TCP connection, not per-packet). You can make of course get it to be per-packet load balanced, but as you note, there are issues with that when you don't control both ends.
  - jerkstate 8y ago
    that really depends on which fields are included in the ecmp hash and can break stuff in weird ways, like path mtu discovery
  - isostatic 8y ago
    You can use a hash of source and dest ip, protocol and port, but you will get confusing results and some sites won't be happy. Using source ip to round robin on active wan connections is the safest.

But, can you toss the connections from both providers into a switch and make both avaialable for use all the time? Like a Active-active setup?? Any reason not to do that?

mansilladev 8y ago

And put my remote hands worker out of a job? No way.

caust1c 8y ago
Yeah, but they must cost you a bundle!
- mansilladev 8y ago
  That, my friend, is a sunk cost.
  - paul_milovanov 8y ago
    See: sunk cost fallacy
  - cardiffspaceman 8y ago
    So you are committed to paying for the remote hands and can't get out of it, or have already fully paid for lifetime support from the remote hands and can't get a refund :) That doesn't mean that there can't be cost savings from automation. For example if it costs X in lost business due to a misunderstanding by the remote hands that extends the outage unnecessarily, then a certain number of times avoiding that X cost would pay for the investment in automation. You pay the "sunk" and the new cost but you avoid unnecessary costs in the long run. It's all a matter of fully modeling your costs and benefits. Noting that certain costs are sunk is a partial model.

TimTheTinker 8y ago

From 5 minutes of Wikipedia reading, it appears this requires special support from the router to enable “sticky sessions” which prevent out-of-order packets; or, the device OS itself can “stripe” packets across NICs with special (extra) software to enable that.

XorNot 8y ago
ECMP handles this just fine by default with Linux (it's per flow aka TCP connection, not per-packet). You can make of course get it to be per-packet load balanced, but as you note, there are issues with that when you don't control both ends.
- jerkstate 8y ago
  that really depends on which fields are included in the ecmp hash and can break stuff in weird ways, like path mtu discovery
- isostatic 8y ago
  You can use a hash of source and dest ip, protocol and port, but you will get confusing results and some sites won't be happy. Using source ip to round robin on active wan connections is the safest.