- mice-discuss - lists.micemn.net

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Mike Horwath

On Fri, Oct 28, 2016 at 09:45:00AM -0500, Jeremy Lumby wrote: > I do believe the result of that discussion is that we will not offer > service level agreements, and that whenever a new member agreement > was drafted, it would essentially say that we cannot guarantee > uptime, or resolution response time, and the solution would be that > the member would responsible to disconnect themselves if > quality/reliability issues arose that the member found unacceptable. That's also just fine. You can have some defined parameters about scheduled work (time, announcements, etc) yet still skip an SLA. -- Mike Horwath, reachable via drechsau(a)Geeks.ORG

Oct. 28, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Mike Horwath

On Thu, Oct 27, 2016 at 02:07:39PM -0500, Doug McIntyre wrote: > On Thu, Oct 27, 2016 at 08:21:54AM -0500, Mike Horwath wrote: > > > EX4550 is 32 ports base, 48 ports with addon cards. > > It is 32 + 8 ports if you want to use the back expansion slot to house > a VC card instead. The 4550 doesn't have built-in VC (unlike the 4200/4500) > (technically the VC card is slotted into the 4500, but I've never seen one > bought without it there). So the EX4500 can do more ports than a EX4550. Ooh I didn't know that! Thanks! > The QFX5200-32C can get into newer technologies like adding 25G/50G > and 100G capable ports. Again using breakout cables to breakout ports > into the "slower" speeds we are running, and can get into the range of > 128 x 10GB ports with breakout cables in 1U. > > The QFX10002-36Q is much like the above, and can get into 144 x 10G > ports with breakout cables, but I know the price on this switch pushes > now into 6 figures. I have to assume that regardless they are expensive. If only that non-profit thing were in place *today*. A lot of work and I don't envy anyone having to do it. But I, and I don't think I am alone, am not donating money until there is a formal organization wrapped up in pretty bows. I have faith, though. This is coming and there will be much celebration. -- Mike Horwath, reachable via drechsau(a)Geeks.ORG

Oct. 28, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Mike Horwath

On Tue, Oct 25, 2016 at 01:56:32PM -0500, Richard Laager wrote: > 3) asking for a MICE-TECH and/or MICE-OPS list be created (depending > on whether we have distinct groups and what those people want). I > expect at least one list will be closed to members only. Give the word and it can be done. Hell both could be done though they seem a little redundant in naming. :) -- Mike Horwath, reachable via drechsau(a)Geeks.ORG

Oct. 28, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Jeremy Lumby

I do believe the result of that discussion is that we will not offer service level agreements, and that whenever a new member agreement was drafted, it would essentially say that we cannot guarantee uptime, or resolution response time, and the solution would be that the member would responsible to disconnect themselves if quality/reliability issues arose that the member found unacceptable. -----Original Message----- From: MICE Discuss [mailto:MICE-DISCUSS@LISTS.IPHOUSE.NET] On Behalf Of Mike Horwath Sent: Friday, October 28, 2016 9:40 AM To: MICE-DISCUSS(a)LISTS.IPHOUSE.NET Subject: Re: [MICE-DISCUSS] MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT On Tue, Oct 25, 2016 at 09:58:06AM -0500, Colin Baker wrote: > I remember some discussion in a MICE UG meeting a few years back that > some larger networks may have strict requirements that must be met in > order to connect to the exchange. We've added some big content > providers recently, which I'm thrilled to see, but I think establishing > some defined and documented standards can only help continue the growth. Confirm and agree. -- Mike Horwath, reachable via drechsau(a)Geeks.ORG

Oct. 28, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Mike Horwath

On Tue, Oct 25, 2016 at 10:10:00AM -0500, Jeremy Lumby wrote: > I agree standards would help, however I know that my personal > donations to MICE would probably have been far less if I was forced > to do it on a corporate standard schedule, and lets remember > donations like that got us here, and I would prefer not to lose > donors over standards, and I do not feel membership will be impacted > because of lack of standards since all of our current members have > gotten along quite well all this time without them. These aren't corporate standard schedules. While I am now on the outside of the service provider world, and no longer a direct member of MICE in any capacity (mailing list doens't count), I do feel that we should have something defined and adhered to for the benefit of the group as a whole. We didn't get here all willy nilly, using 1994 ISP style operations. Should there be defined announcement windows for scheduled maintenance? I think most everyone would agree that would be a yes. Should there be notifications of unscheduled outages, maintenance, lunch changes? Yes but this would be looser for timeframes. In the heat of the moment would you rather see work done to further forward the repair or would you like notifications every 15 minutes that things are still broken? Tough stuff.. -- Mike Horwath, reachable via drechsau(a)Geeks.ORG

Oct. 28, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Mike Horwath

On Tue, Oct 25, 2016 at 09:58:06AM -0500, Colin Baker wrote: > I remember some discussion in a MICE UG meeting a few years back that > some larger networks may have strict requirements that must be met in > order to connect to the exchange. We've added some big content > providers recently, which I'm thrilled to see, but I think establishing > some defined and documented standards can only help continue the growth. Confirm and agree. -- Mike Horwath, reachable via drechsau(a)Geeks.ORG

Oct. 28, 2016

BGP with BFD support

by Doug McIntyre

We had been testing BFD support for BGP in the last few days, and I specificly watched it during the maintenance last night. I currently have route server 1 configured for passive BFD for a couple members, and that has been working just fine as far as I can tell for myself. I'm ready to open it up to any member that wants to test, just let me know your ASN and that you want passive BFD setup, and I can get that going on route-server 1. I'll do IPv4 and IPv6 sessions both. I assume everybody wants more of a gradual test. Otherwise I could enable it on route-server 2 as well. Finally, since this is a passive setting until triggered.. Should this get applied for every member by default? As long as the member side doesn't trigger it, it sits dormant. I could see that being part of the default template rollout for everybody, as well as every member retroactively, so it is totally controlled by what the member wants to do. If they want it, they do it. If they don't care to, it just sits passive not doing it. -- Doug McIntyre <merlyn(a)iphouse.net> ~.~ ipHouse ~.~ Network Engineer/Provisioning/Jack of all Trades

Oct. 27, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by David Farmer

On Thu, Oct 27, 2016 at 2:07 PM, Doug McIntyre <merlyn(a)iphouse.net> wrote: > On Thu, Oct 27, 2016 at 08:21:54AM -0500, Mike Horwath wrote: > > On Thu, Oct 27, 2016 at 08:06:45AM -0500, Jason Hanke wrote: > > > There aren't enough ports on the 4550 to swap it in we would need > > > two. We are due for a fabric upgrade. > > Correct, while one could be brought in with additional cards in both, > that would barely cover us with zero future growth without adding > in a 2nd one. The U of MN expressed interest in a 40Gbps connection, > and doing just one switch would knock that out of contention, (or > we keep the EX4500 and run it knowing that it is EOL, and software will > probably stop being available within 14 months). > Don't worry about a 40G for us I'm looking past that now to 100G. I think the priority is to stabilize what we have and add a 100G/50G/40G/25G capability soon next two or three months. Thanks. -- =============================================== David Farmer Email:farmer@umn.edu Networking & Telecommunication Services Office of Information Technology University of Minnesota 2218 University Ave SE Phone: 612-626-0815 Minneapolis, MN 55414-3029 Cell: 612-812-9952 ===============================================

Oct. 27, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Levi Pederson

Doug, We've just finished some research into 10Gig + 40 to 100Gig Density mapping in Juniper. I'll collect my notes and send them to the list with my thoughts. Thank you, *Levi Pederson* Mankato Networks LLC cell | 612.481.0769 work | 612.787.7392 levipederson(a)mankatonetworks.net On Thu, Oct 27, 2016 at 2:07 PM, Doug McIntyre <merlyn(a)iphouse.net> wrote: > On Thu, Oct 27, 2016 at 08:21:54AM -0500, Mike Horwath wrote: > > On Thu, Oct 27, 2016 at 08:06:45AM -0500, Jason Hanke wrote: > > > There aren't enough ports on the 4550 to swap it in we would need > > > two. We are due for a fabric upgrade. > > Correct, while one could be brought in with additional cards in both, > that would barely cover us with zero future growth without adding > in a 2nd one. The U of MN expressed interest in a 40Gbps connection, > and doing just one switch would knock that out of contention, (or > we keep the EX4500 and run it knowing that it is EOL, and software will > probably stop being available within 14 months). > > > EX4550 is 32 ports base, 48 ports with addon cards. > > It is 32 + 8 ports if you want to use the back expansion slot to house > a VC card instead. The 4550 doesn't have built-in VC (unlike the 4200/4500) > (technically the VC card is slotted into the 4500, but I've never seen one > bought without it there). So the EX4500 can do more ports than a EX4550. > > So, the back slot on a 4550 could be another 8x10G, but that loses the > VC port, which means we then have to take multiples of 10G ports on > both switches to devote to VC instead, probably at least 4 x 10GB on > both switches, which is a losing of 8 ports either way. > > Effectively making the stacked EX4550 only 40 ports instead of the 48 ports > on the 4500. > > > Is the current EX4500 running with 40 ports or 48? > > It has 48 ports, with 47 ports plugged in. (didn't count how many as lit, > but we just did an inventory and I know how few empty ports we have). > > The EX4550 is the direct replacement for the EX4500. I've run multiple > stacks of them myself, and they are solid boxes. > > Juniper also has newer switches at higher price points (and not as > easy upgrade path). The EX4600 can do 72 x 10GB in 1U. It has 40 fixed > 10G, 4 x 40G. BUT no VC ports. You use the 40G ports for VC > functions. To get the 72 port density, you'd add in a 14 x 40G card, > and any and each 40G port can be broken out to 4 x 10G port with breakout > cables. So, stacking the EX4600's typically takes out the fixed 4x40G > ports with the QSFP+ DAC cabling up and down, but adding on the 14 x 40G > card > brings back many ports and capabilities. > The EX4600 won't VC with the EX4550 so it becomes another stack connected > in. > > The QFX5200-32C can get into newer technologies like adding 25G/50G > and 100G capable ports. Again using breakout cables to breakout ports > into the "slower" speeds we are running, and can get into the range of > 128 x 10GB ports with breakout cables in 1U. > > The QFX10002-36Q is much like the above, and can get into 144 x 10G > ports with breakout cables, but I know the price on this switch pushes > now into 6 figures. > > > > > -- > Doug McIntyre <merlyn(a)iphouse.net> > ~.~ ipHouse ~.~ > Network Engineer/Provisioning/Jack of all Trades >

Oct. 27, 2016

Re: MICE core switch maintenance window - 10/27/2016 12:30AM-1:00AM CDT

by Doug McIntyre

On Thu, Oct 27, 2016 at 08:21:54AM -0500, Mike Horwath wrote: > On Thu, Oct 27, 2016 at 08:06:45AM -0500, Jason Hanke wrote: > > There aren't enough ports on the 4550 to swap it in we would need > > two. We are due for a fabric upgrade. Correct, while one could be brought in with additional cards in both, that would barely cover us with zero future growth without adding in a 2nd one. The U of MN expressed interest in a 40Gbps connection, and doing just one switch would knock that out of contention, (or we keep the EX4500 and run it knowing that it is EOL, and software will probably stop being available within 14 months). > EX4550 is 32 ports base, 48 ports with addon cards. It is 32 + 8 ports if you want to use the back expansion slot to house a VC card instead. The 4550 doesn't have built-in VC (unlike the 4200/4500) (technically the VC card is slotted into the 4500, but I've never seen one bought without it there). So the EX4500 can do more ports than a EX4550. So, the back slot on a 4550 could be another 8x10G, but that loses the VC port, which means we then have to take multiples of 10G ports on both switches to devote to VC instead, probably at least 4 x 10GB on both switches, which is a losing of 8 ports either way. Effectively making the stacked EX4550 only 40 ports instead of the 48 ports on the 4500. > Is the current EX4500 running with 40 ports or 48? It has 48 ports, with 47 ports plugged in. (didn't count how many as lit, but we just did an inventory and I know how few empty ports we have). The EX4550 is the direct replacement for the EX4500. I've run multiple stacks of them myself, and they are solid boxes. Juniper also has newer switches at higher price points (and not as easy upgrade path). The EX4600 can do 72 x 10GB in 1U. It has 40 fixed 10G, 4 x 40G. BUT no VC ports. You use the 40G ports for VC functions. To get the 72 port density, you'd add in a 14 x 40G card, and any and each 40G port can be broken out to 4 x 10G port with breakout cables. So, stacking the EX4600's typically takes out the fixed 4x40G ports with the QSFP+ DAC cabling up and down, but adding on the 14 x 40G card brings back many ports and capabilities. The EX4600 won't VC with the EX4550 so it becomes another stack connected in. The QFX5200-32C can get into newer technologies like adding 25G/50G and 100G capable ports. Again using breakout cables to breakout ports into the "slower" speeds we are running, and can get into the range of 128 x 10GB ports with breakout cables in 1U. The QFX10002-36Q is much like the above, and can get into 144 x 10G ports with breakout cables, but I know the price on this switch pushes now into 6 figures. -- Doug McIntyre <merlyn(a)iphouse.net> ~.~ ipHouse ~.~ Network Engineer/Provisioning/Jack of all Trades

Oct. 27, 2016