Arista vs Nexus, which would you rollout?

[burnyd]

Haha just ISSU in general.  Buy small rack unit switches run them a as spine switches and scale out your blast radius.

[wintermute000]

+100
I still have nightmares about my last multi stage (3 step) ISSU project for 7ks.

[NetworkGroover]

ISSU  is the lols

ISSU with single processor devices

 

SSU Leaf - challenge accepted.

[NetworkGroover]

And in regards to ISSU, I did recently sit on a 2am to 5 am change control window where three MLAG ISSU upgrades were performed - went off without a hitch - I was pleasantly shocked.

[NetworkGroover]

At the end of the day, however, accounting is architecture. If you want to justify actually spending money on hardware vs. free hardware, you've got a lot of homework ahead of you. Cisco knows this, which is why they cut way back on their purchase prices, hoping to make the money back in licensing renewals and professional services engagements. In accounting-speak, this is all part of TCO, total cost of ownership.

If you can put numbers on *all* the costs and show that vendor A is going to be cheaper than vendor C, J, or even B, then you'll have numbers the accountants can understand. Any savings on personnel costs, maintenance, licensing, and the like will be important to identify and include.

Think of it this way: if we spend $400K on gear I can configure myself, that's ultimately cheaper than free gear that requires a $500K 6-month pro serv engagement with the vendor in order to get up and running. If you estimate more vendor staff would be required, say an actual on-site, year-round technical resource, that can easily hit a million, depending upon the vendor. That million may include all-hours tech support and a case manager, but that's still way more than the gear that has a higher cost that you can set up on your own with a minimum of tech calls.

I have a hard time swallowing this pill, but I haven't worked a day in an operational role.  For me, I'm very much the type of person to go against the grain and be like, "Look - this is the best solution for XYZ reasons." (I also am completely capable of being able to tell when a solution is overkill as well, so when I say "best", I mean it's the best fit - cost is a secondary concern for me).  I come at it from a no-nonsense, simple approach.  If they can't stomach the price tag, I simply ask what that price needs to be, then I go back to the vendor.  If they can't do it, I find the next best solution.  If I start approaching a deadline, that's not my fault - I did my job.  I presented the best solution possible, and you chose not to accept it and have me find another.

I'd probably be fired within a year.

[deanwebb]

I'd probably be fired within a year.

Probably, yes. 

But what you're saying about "next best solution"... these guys will go for a totally unworkable solution, if the price is right and the vendor does enough of a C-level end-run and character assassination of an engineer to discount what's being said against their product. 

I can say that A and B work great, C is abomination - use it and die.

C can then drop its pants on pricing, show a bunch of Gartner slides to the managers, and snow the higher-ups to where they ignore my recommendation and go with C because they *have* to have a vendor for this solution and, besides, nobody ever got fired for buying C, right?

Not only do I know of guys who got fired for buying C, I also know people that left companies after they bought C because they knew what was coming next.

[wintermute000]

The big C is introducing software-ISSU - containers LOL - other vendors doing similar. Though in this day and age, plan for failure, ECMP leaf-spine, just take it out gracefully, upgrade and bring it back into routing, done.

[LynK]

yeah but how gracefully can you take out a spine switch running L3, and have it be stateful? There is no way, even with BFD which makes it faster, but not stateful.

Thoughts?

[wintermute000]

http://www.cisco.com/c/en/us/products/collateral/switches/nexus-7000-series-switches/white-paper-c11-735628.html#_Toc427398694

[wintermute000]

Also graceful bgp shutdown is a thing but haven't checked 9k support. Just normal routing things...

If theres a way to do it with standard routing, it's the answer.



So yeah, there is most definitely a way, and people do it all the time. OSPF max-metric is an old ISP trick for OSPF/iBGP MPLS designs for example, tweaking metrics is foolproof and yeah hitless - you may have some reordered packets but TCP should handle it

[LynK]

Thanks for the article. I was aware of graceful BGP, but not of the others.

http://www.cisco.com/c/en/us/td/docs/switches/datacenter/nexus9000/sw/7-x/system_management/configuration/guide/b_Cisco_Nexus_9000_Series_NX-OS_System_Management_Configuration_Guide_7x/b_Cisco_Nexus_9000_Series_NX-OS_System_Management_Configuration_Guide_7x_chapter_011010.html

[NetworkGroover]

yeah but how gracefully can you take out a spine switch running L3, and have it be stateful? There is no way, even with BFD which makes it faster, but not stateful.

Thoughts?

Poison metrics, or:

BGP Maintenance Mode:
https://eos.arista.com/maintenance-mode-lab-example-of-bgp-on-spine/

[NetworkGroover]

I'd probably be fired within a year.

Probably, yes. 

But what you're saying about "next best solution"... these guys will go for a totally unworkable solution, if the price is right and the vendor does enough of a C-level end-run and character assassination of an engineer to discount what's being said against their product. 

I can say that A and B work great, C is abomination - use it and die.

C can then drop its pants on pricing, show a bunch of Gartner slides to the managers, and snow the higher-ups to where they ignore my recommendation and go with C because they *have* to have a vendor for this solution and, besides, nobody ever got fired for buying C, right?

Not only do I know of guys who got fired for buying C, I also know people that left companies after they bought C because they knew what was coming next.

Yep - I'd be fired or quit and find a job elsewhere.  Frankly, I don't put up with that crap and am completely confident I can find employment elsewhere that values me as an employee.

EDIT - Isn't this easily addressed though by requesting/demanding to be part of any conversations that involve infrastructure, since you're the one in charge/has to maintain/is the throat that gets choked when it breaks?  I mean, that's completely reasonable, right?  That way when they try to sling FUD you can address it right on the spot?

[mlan]

Isn't this easily addressed though by requesting/demanding to be part of any conversations that involve infrastructure, since you're the one in charge/has to maintain/is the throat that gets choked when it breaks?  I mean, that's completely reasonable, right?  That way when they try to sling FUD you can address it right on the spot?

Your thoughts are completely reasonable.  The problem in a lot of orgs (both private and public) is that business culture and politics drive a lot of these big "we choose vendor X" decisions, and unless you are willing to drive culture and play politics you could be left out of the decision.  It's a lot to take on for sr. engineers that might want to avoid such things.  It's also a great way to get a potentially unwanted promotion into leadership/management, haha.  Throw in public sector RFP requirements and it's a whole other can of worms.  That said, I would venture that the relationship between the sr. engineers and their  manager/CTO/CIO is the crux of the matter.  If there is trust and respect there, then it's probably a non-issue, but if there is no trust... 

[that1guy15]

@ AspiringNetworker

Oh the number of times I have faced any of the following or similar.

- CEO sat next to another CEO in 1st one trip. Determined we will now deploy $new_hotness from $vendor. "Signing contract today"
- Apps team with large project budget purchases network with consultants to deploy EVERYTHING. Its next week. "Hey That1guy you got a couple min to talk through how this will connect?"
- Submitted network design/BOM/PO rejected. Leadership and procurement rejected due to "$crappySwitchCo undercut bit" and this is what was purchased. "Can you have it deployed in 2 weeks?"

Its sad but I see more times project design goes as far as what switches to purchase before cutting a PO. Rest can be figured out after we get them in.