Login | Register For Free | Help
Search for: (Advanced)

Mailing List Archive: Cisco: NSP

Stacking 3750X vs diverse 4948E

  

 
First page Previous page 1 2 Next page Last page  View All Cisco nsp RSS feed   Index | Next | Previous | View Threaded


david at davidcoulson

May 18, 2012, 11:55 AM

Post #1 of 40 (2880 views)
Permalink
Stacking 3750X vs diverse 4948E
In a datacenter environment, we typically deploy 4948 top-of-rack
switches with L2 uplinks to our 6500 core - Systems get connections into
two different switches and rely on OS NIC bonding (mostly Linux) to
support switch failures. Switches running STP and in the last four years
we've had no issues with this design (including failures of systems
connected to diverse switches).

A new proposed configuration utilizes stacked 3750X switches, where
servers would be connected to multiple switches within the same stack. I
have next to no experience in the low-end switches that do stacking, but
from a general risk management perspective, it seems like a many eggs
and single basket configuration.

Does anyone have any solid experience with 3750X switches, or stacking
in a datacenter in general? I've seen plenty of stacks for
closets/end-users, but I don't see many in a top-of-rack config. Is
Cisco stacking typically 'reliable', in that when a switch fails it will
leave the remainder of the stack functional? What about a software
issue? Does the whole stack crap out and reload, or does the master just
fail and a new one get elected?

I realize it's a pretty broad question, but it boils down to - Is a
stacked switch config significantly less reliable/resilient/available
than two TOR switches?

David

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


keegan.holley at sungard

May 18, 2012, 2:00 PM

Post #2 of 40 (2819 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
The 3750X is relatively new so I've only seen a few of them. Stackwise in
general is pretty solid. I've never seen a whole stack fail. If a member
fails the stack just keeps going, if the master tails a new master is
elected. One thing to watch out for is the fact that the 3750X isn't
intended to be a high performance DC switch. I have seen issues with queue
drops because of small packet buffers on the non-X version which leads to
trouble if you do alot of 1G at line rate. I haven't checked the X series,
but I'm told it's not recommended for high performance environments either.


2012/5/18 David Coulson <david [at] davidcoulson>

> In a datacenter environment, we typically deploy 4948 top-of-rack switches
> with L2 uplinks to our 6500 core - Systems get connections into two
> different switches and rely on OS NIC bonding (mostly Linux) to support
> switch failures. Switches running STP and in the last four years we've had
> no issues with this design (including failures of systems connected to
> diverse switches).
>
> A new proposed configuration utilizes stacked 3750X switches, where
> servers would be connected to multiple switches within the same stack. I
> have next to no experience in the low-end switches that do stacking, but
> from a general risk management perspective, it seems like a many eggs and
> single basket configuration.
>
> Does anyone have any solid experience with 3750X switches, or stacking in
> a datacenter in general? I've seen plenty of stacks for closets/end-users,
> but I don't see many in a top-of-rack config. Is Cisco stacking typically
> 'reliable', in that when a switch fails it will leave the remainder of the
> stack functional? What about a software issue? Does the whole stack crap
> out and reload, or does the master just fail and a new one get elected?
>
> I realize it's a pretty broad question, but it boils down to - Is a
> stacked switch config significantly less reliable/resilient/available than
> two TOR switches?
>
> David
>
> ______________________________**_________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/**mailman/listinfo/cisco-nsp<https://puck.nether.net/mailman/listinfo/cisco-nsp>
> archive at http://puck.nether.net/**pipermail/cisco-nsp/<http://puck.nether.net/pipermail/cisco-nsp/>
>
>
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


nsp.alexander.lim at gmail

May 19, 2012, 1:11 AM

Post #3 of 40 (2807 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
When doing IOS upgrade, you need to reboot the whole switches in the stack.

Regards,
Alexander Halim

On May 19, 2012, at 5:00 AM, Keegan Holley <keegan.holley [at] sungard> wrote:

> The 3750X is relatively new so I've only seen a few of them. Stackwise in
> general is pretty solid. I've never seen a whole stack fail. If a member
> fails the stack just keeps going, if the master tails a new master is
> elected. One thing to watch out for is the fact that the 3750X isn't
> intended to be a high performance DC switch. I have seen issues with queue
> drops because of small packet buffers on the non-X version which leads to
> trouble if you do alot of 1G at line rate. I haven't checked the X series,
> but I'm told it's not recommended for high performance environments either.
>
>
> 2012/5/18 David Coulson <david [at] davidcoulson>
>
>> In a datacenter environment, we typically deploy 4948 top-of-rack switches
>> with L2 uplinks to our 6500 core - Systems get connections into two
>> different switches and rely on OS NIC bonding (mostly Linux) to support
>> switch failures. Switches running STP and in the last four years we've had
>> no issues with this design (including failures of systems connected to
>> diverse switches).
>>
>> A new proposed configuration utilizes stacked 3750X switches, where
>> servers would be connected to multiple switches within the same stack. I
>> have next to no experience in the low-end switches that do stacking, but
>> from a general risk management perspective, it seems like a many eggs and
>> single basket configuration.
>>
>> Does anyone have any solid experience with 3750X switches, or stacking in
>> a datacenter in general? I've seen plenty of stacks for closets/end-users,
>> but I don't see many in a top-of-rack config. Is Cisco stacking typically
>> 'reliable', in that when a switch fails it will leave the remainder of the
>> stack functional? What about a software issue? Does the whole stack crap
>> out and reload, or does the master just fail and a new one get elected?
>>
>> I realize it's a pretty broad question, but it boils down to - Is a
>> stacked switch config significantly less reliable/resilient/available than
>> two TOR switches?
>>
>> David
>>
>> ______________________________**_________________
>> cisco-nsp mailing list cisco-nsp [at] puck
>> https://puck.nether.net/**mailman/listinfo/cisco-nsp<https://puck.nether.net/mailman/listinfo/cisco-nsp>
>> archive at http://puck.nether.net/**pipermail/cisco-nsp/<http://puck.nether.net/pipermail/cisco-nsp/>
>>
>>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


nick at foobar

May 19, 2012, 3:32 AM

Post #4 of 40 (2805 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 19/05/2012 09:11, Alexander Lim wrote:
> When doing IOS upgrade, you need to reboot the whole switches in the stack.

yes - and this can cause unexpectedly long outages during your maintenance
windows. I've found the stacking to be very reliable in general, but the
upgrade hit is bad news particularly if the code upgrade includes microcode
and bootloader upgrades:

http://puck.nether.net/pipermail/cisco-nsp/2010-December/075697.html

Because of the hit associated with this, I've destacked several existing
stacks and am not planning to use this feature in future. Your network
requirements may be different though.

Nick


> Regards,
> Alexander Halim
>
> On May 19, 2012, at 5:00 AM, Keegan Holley <keegan.holley [at] sungard> wrote:
>
>> The 3750X is relatively new so I've only seen a few of them. Stackwise in
>> general is pretty solid. I've never seen a whole stack fail. If a member
>> fails the stack just keeps going, if the master tails a new master is
>> elected. One thing to watch out for is the fact that the 3750X isn't
>> intended to be a high performance DC switch. I have seen issues with queue
>> drops because of small packet buffers on the non-X version which leads to
>> trouble if you do alot of 1G at line rate. I haven't checked the X series,
>> but I'm told it's not recommended for high performance environments either.
>>
>>
>> 2012/5/18 David Coulson <david [at] davidcoulson>
>>
>>> In a datacenter environment, we typically deploy 4948 top-of-rack switches
>>> with L2 uplinks to our 6500 core - Systems get connections into two
>>> different switches and rely on OS NIC bonding (mostly Linux) to support
>>> switch failures. Switches running STP and in the last four years we've had
>>> no issues with this design (including failures of systems connected to
>>> diverse switches).
>>>
>>> A new proposed configuration utilizes stacked 3750X switches, where
>>> servers would be connected to multiple switches within the same stack. I
>>> have next to no experience in the low-end switches that do stacking, but
>>> from a general risk management perspective, it seems like a many eggs and
>>> single basket configuration.
>>>
>>> Does anyone have any solid experience with 3750X switches, or stacking in
>>> a datacenter in general? I've seen plenty of stacks for closets/end-users,
>>> but I don't see many in a top-of-rack config. Is Cisco stacking typically
>>> 'reliable', in that when a switch fails it will leave the remainder of the
>>> stack functional? What about a software issue? Does the whole stack crap
>>> out and reload, or does the master just fail and a new one get elected?
>>>
>>> I realize it's a pretty broad question, but it boils down to - Is a
>>> stacked switch config significantly less reliable/resilient/available than
>>> two TOR switches?
>>>
>>> David
>>>
>>> ______________________________**_________________
>>> cisco-nsp mailing list cisco-nsp [at] puck
>>> https://puck.nether.net/**mailman/listinfo/cisco-nsp<https://puck.nether.net/mailman/listinfo/cisco-nsp>
>>> archive at http://puck.nether.net/**pipermail/cisco-nsp/<http://puck.nether.net/pipermail/cisco-nsp/>
>>>
>>>
>> _______________________________________________
>> cisco-nsp mailing list cisco-nsp [at] puck
>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


daniel at fnutt

May 19, 2012, 3:41 AM

Post #5 of 40 (2801 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 19/5/12 10:11 , Alexander Lim wrote:
> When doing IOS upgrade, you need to reboot the whole switches in the stack.

No, in 12.2(58) 3750E and X got support for RSU (rolling stack upgrade).

--
Daniel
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


saku at ytti

May 19, 2012, 4:21 AM

Post #6 of 40 (2794 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On (2012-05-18 14:55 -0400), David Coulson wrote:

> Does anyone have any solid experience with 3750X switches, or
> stacking in a datacenter in general? I've seen plenty of stacks for

We've had quite many 3750 stacks, and we do see more problems in them than
in other 3750 stacks.
Particularly dangerous events are adding or removing members from stack.

Also we've seen software defects which only affect stacks, like memory
leakage on SNMP polling.

Generally, do not add software complexity to the network if you have any
practical alternative. STP while PITA, is tried and true and it actually
does work in CSCO when configured right. So unless you absolutely need the
extra interconnect capacity, go with STP.

--
++ytti
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


ler762 at gmail

May 19, 2012, 4:47 AM

Post #7 of 40 (2793 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/19/12, Saku Ytti <saku [at] ytti> wrote:
> On (2012-05-18 14:55 -0400), David Coulson wrote:
>
>> Does anyone have any solid experience with 3750X switches, or
>> stacking in a datacenter in general? I've seen plenty of stacks for
>
> We've had quite many 3750 stacks, and we do see more problems in them than
> in other 3750 stacks.
> Particularly dangerous events are adding or removing members from stack.
>
> Also we've seen software defects which only affect stacks, like memory
> leakage on SNMP polling.
>
> Generally, do not add software complexity to the network if you have any
> practical alternative. STP while PITA, is tried and true and it actually
> does work in CSCO when configured right. So unless you absolutely need the
> extra interconnect capacity, go with STP.

How about VSS? We're considering it mainly because it would eliminate STP

Thanks,
Lee
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


saku at ytti

May 19, 2012, 5:10 AM

Post #8 of 40 (2793 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On (2012-05-19 07:47 -0400), Lee wrote:

> How about VSS? We're considering it mainly because it would eliminate STP

There are already horror stories in c-nsp, where software defect has taken
whole VSS cluster down. STP is very unlikely to do that, as the code is lot
simpler and lot more mature.


To me it is clear that main things that cause outages are

1. Operator
2. Software defect
3. Hardware defect

And there are huge gap between probability of each, i.e. operator is much
more likely to break the network than software defect, and so forth.

Yet typically even high budget, high clue, critical importance networks are
designed with only working around outages caused by 3. Often these efforts
actually increase probability of 1 and 2. Essentially often the 'well
design' network has lower MTBF due to the added software complexity.

Key example here is stateful firewall clusters, which I consistently see
failing more often than single firewalls.
When possible, I would separate elements with routing and accept that users
will see sessions breaking when there is network fault.

If you keep eye open in press, these examples are on the news all the time,
where CIO explains that the setup was fully redundant yadayada, it should
have never failed.

Latest example I can think of was large outsourcing/integrator losing their
whole 'redundant' storage setup, causing 5 day outage. Or bit longer ago,
public sector health care had to resort to dead wood as LAN was down for
1.5weeks.
Both were designed not to fail and neither was designed to workaround (or
even rapid recover from) software defects.

--
++ytti
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


mkorourke at gmail

May 19, 2012, 5:18 AM

Post #9 of 40 (2801 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
> The 3750X is relatively new so I've only seen a few of them. Stackwise in
> general is pretty solid. I've never seen a whole stack fail. If a member
> fails the stack just keeps going, if the master tails a new master is
> elected.


For the most part my experience is in-line with the above, yet when they do
fail they fail in a bad way and you can get some very strange results
across the entire stack - a good example has been with cross stack
etherchannel.

All said you don't have to stack them if your requirement is ultra mission
critical and if the tcam sdm templates fit your requirements then they're
not a bad layer2\3 switch IMHO. Vrf-lite support, dual hot swap power, hot
swap fans. A 4500\4900 without the tcam and minus some software feature
support. They'll likely do the job.

Mick
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


scottowens12 at gmail

May 19, 2012, 5:03 PM

Post #10 of 40 (2781 views)
Permalink
Stacking 3750X vs diverse 4948E [In reply to]
How about Nexus 5010s.

I think they bundle them for less than 2 x 3750X .
We have both but the 3750s are used where we needed L2/L3, the 5Ks for just
L2 up to VSS or 7Ks.


you can boot them separately and they do LACP / Etherchannel just fine.




> 2. Stacking 3750X vs diverse 4948E (David Coulson)
> Message: 2
> Date: Fri, 18 May 2012 14:55:57 -0400
> From: David Coulson <david [at] davidcoulson>
> To: cisco-nsp [at] puck
> Subject: [c-nsp] Stacking 3750X vs diverse 4948E
> Message-ID: <4FB69B3D.3060802 [at] davidcoulson>
> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>
> In a datacenter environment, we typically deploy 4948 top-of-rack
> switches with L2 uplinks to our 6500 core - Systems get connections into
> two different switches and rely on OS NIC bonding (mostly Linux) to
> support switch failures. Switches running STP and in the last four years
> we've had no issues with this design (including failures of systems
> connected to diverse switches).
>
> A new proposed configuration utilizes stacked 3750X switches, where
> servers would be connected to multiple switches within the same stack. I
> have next to no experience in the low-end switches that do stacking, but
> from a general risk management perspective, it seems like a many eggs
> and single basket configuration.
>
> Does anyone have any solid experience with 3750X switches, or stacking
> in a datacenter in general? I've seen plenty of stacks for
> closets/end-users, but I don't see many in a top-of-rack config. Is
> Cisco stacking typically 'reliable', in that when a switch fails it will
> leave the remainder of the stack functional? What about a software
> issue? Does the whole stack crap out and reload, or does the master just
> fail and a new one get elected?
>
> I realize it's a pretty broad question, but it boils down to - Is a
> stacked switch config significantly less reliable/resilient/available
> than two TOR switches?
>
> David
>
>
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


nsp.alexander.lim at gmail

May 19, 2012, 7:47 PM

Post #11 of 40 (2788 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
Thanks for the correction.

Regards,
Alexander Lim

On May 19, 2012, at 6:41 PM, Daniel Husand <daniel [at] fnutt> wrote:

> On 19/5/12 10:11 , Alexander Lim wrote:
>> When doing IOS upgrade, you need to reboot the whole switches in the stack.
>
> No, in 12.2(58) 3750E and X got support for RSU (rolling stack upgrade).
>
> --
> Daniel

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


nsp.alexander.lim at gmail

May 19, 2012, 7:48 PM

Post #12 of 40 (2780 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
Wouw....scary.
Any clue if Nexus is better than VSS?

Regards,
Alexander Lim

On May 19, 2012, at 8:10 PM, Saku Ytti <saku [at] ytti> wrote:

> On (2012-05-19 07:47 -0400), Lee wrote:
>
>> How about VSS? We're considering it mainly because it would eliminate STP
>
> There are already horror stories in c-nsp, where software defect has taken
> whole VSS cluster down. STP is very unlikely to do that, as the code is lot
> simpler and lot more mature.
>
>
> To me it is clear that main things that cause outages are
>
> 1. Operator
> 2. Software defect
> 3. Hardware defect
>
> And there are huge gap between probability of each, i.e. operator is much
> more likely to break the network than software defect, and so forth.
>
> Yet typically even high budget, high clue, critical importance networks are
> designed with only working around outages caused by 3. Often these efforts
> actually increase probability of 1 and 2. Essentially often the 'well
> design' network has lower MTBF due to the added software complexity.
>
> Key example here is stateful firewall clusters, which I consistently see
> failing more often than single firewalls.
> When possible, I would separate elements with routing and accept that users
> will see sessions breaking when there is network fault.
>
> If you keep eye open in press, these examples are on the news all the time,
> where CIO explains that the setup was fully redundant yadayada, it should
> have never failed.
>
> Latest example I can think of was large outsourcing/integrator losing their
> whole 'redundant' storage setup, causing 5 day outage. Or bit longer ago,
> public sector health care had to resort to dead wood as LAN was down for
> 1.5weeks.
> Both were designed not to fail and neither was designed to workaround (or
> even rapid recover from) software defects.
>
> --
> ++ytti
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


tvarriale at comcast

May 19, 2012, 8:24 PM

Post #13 of 40 (2773 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/18/2012 1:55 PM, David Coulson wrote:
> In a datacenter environment, we typically deploy 4948 top-of-rack
> switches with L2 uplinks to our 6500 core - Systems get connections
> into two different switches and rely on OS NIC bonding (mostly Linux)
> to support switch failures. Switches running STP and in the last four
> years we've had no issues with this design (including failures of
> systems connected to diverse switches).
>
> A new proposed configuration utilizes stacked 3750X switches, where
> servers would be connected to multiple switches within the same stack.
> I have next to no experience in the low-end switches that do stacking,
> but from a general risk management perspective, it seems like a many
> eggs and single basket configuration.
>
> Does anyone have any solid experience with 3750X switches, or stacking
> in a datacenter in general? I've seen plenty of stacks for
> closets/end-users, but I don't see many in a top-of-rack config. Is
> Cisco stacking typically 'reliable', in that when a switch fails it
> will leave the remainder of the stack functional? What about a
> software issue? Does the whole stack crap out and reload, or does the
> master just fail and a new one get elected?
>
> I realize it's a pretty broad question, but it boils down to - Is a
> stacked switch config significantly less reliable/resilient/available
> than two TOR switches?
>
> David
>

The first and most important question is: Is this a real datacenter? If
so, 3750xyz is not for you.

Also, the regular 4948s are not much better.

tv

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


tvarriale at comcast

May 19, 2012, 8:25 PM

Post #14 of 40 (2780 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/19/2012 6:21 AM, Saku Ytti wrote:
> On (2012-05-18 14:55 -0400), David Coulson wrote:
>
>> Does anyone have any solid experience with 3750X switches, or
>> stacking in a datacenter in general? I've seen plenty of stacks for
> We've had quite many 3750 stacks, and we do see more problems in them than
> in other 3750 stacks.
> Particularly dangerous events are adding or removing members from stack.
>
>
If you follow the rules, those are the easiest, most non-eventful events
ever. I've done over 100 and had no issues.

tv
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


tvarriale at comcast

May 19, 2012, 8:26 PM

Post #15 of 40 (2776 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/19/2012 6:47 AM, Lee wrote:
> On 5/19/12, Saku Ytti<saku [at] ytti> wrote:
>> On (2012-05-18 14:55 -0400), David Coulson wrote:
>>
>>> Does anyone have any solid experience with 3750X switches, or
>>> stacking in a datacenter in general? I've seen plenty of stacks for
>> We've had quite many 3750 stacks, and we do see more problems in them than
>> in other 3750 stacks.
>> Particularly dangerous events are adding or removing members from stack.
>>
>> Also we've seen software defects which only affect stacks, like memory
>> leakage on SNMP polling.
>>
>> Generally, do not add software complexity to the network if you have any
>> practical alternative. STP while PITA, is tried and true and it actually
>> does work in CSCO when configured right. So unless you absolutely need the
>> extra interconnect capacity, go with STP.
> How about VSS? We're considering it mainly because it would eliminate STP
VSS doesn't eliminate STP. Neither does vPC. So, you may want to
reconsider.

tv
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


tvarriale at comcast

May 19, 2012, 8:29 PM

Post #16 of 40 (2766 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/19/2012 7:03 PM, scott owens wrote:
> How about Nexus 5010s.
^ +1000000000 other than a missing odd feature. The Nexus 55xx are
purty nice boxen and have HA features that the 375x only dream about.

tv
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


brett at looney

May 19, 2012, 8:33 PM

Post #17 of 40 (2781 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
> Any clue if Nexus is better than VSS?

vPC is not necessarily better than VSS, just different. Be aware that there
are a large number of design caveats around both and you're best to read the
documentation (and blogs and everything else) before making an informed
decision. Having been on the receiving end of a number of those caveats in
the vPC space, my personal preference (if the design allows for it) is VSS
but YMMV. There are cases where VSS simply isn't possible (i.e. you need
features in the Nexus space) and so you have to work with what you get.

If you meet someone who tells you that vPC is simple, walk away and find
someone who knows that they're talking about.

B.

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


skeeve+cisconsp at eintellego

May 19, 2012, 9:54 PM

Post #18 of 40 (2779 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
Feature / Nexus 5010 / 3750X

VLANs / 507 / 1005
MAC / 16k / 4k-12k
L3 / N / Y
vPC / Y / N

Nexus 5010 - less VLANs, no Layer 3, vPC
3750X - more VLAN, Layer 3, no vPC


*Skeeve Stevens, CEO*
eintellego Pty Ltd
skeeve [at] eintellego ; www.eintellego.net <http://www.eintellego.net.au>

Phone: 1300 753 383 ; Fax: (+612) 8572 9954

Cell +61 (0)414 753 383 ; skype://skeeve

facebook.com/eintellego

twitter.com/networkceoau ; www.linkedin.com/in/skeeve

PO Box 7726, Baulkham Hills, NSW 1755 Australia

The Experts Who The Experts Call
Juniper - Cisco – IBM



On Sun, May 20, 2012 at 10:03 AM, scott owens <scottowens12 [at] gmail>wrote:

> How about Nexus 5010s.
>
> I think they bundle them for less than 2 x 3750X .
> We have both but the 3750s are used where we needed L2/L3, the 5Ks for just
> L2 up to VSS or 7Ks.
>
>
> you can boot them separately and they do LACP / Etherchannel just fine.
>
>
>
>
> > 2. Stacking 3750X vs diverse 4948E (David Coulson)
> > Message: 2
> > Date: Fri, 18 May 2012 14:55:57 -0400
> > From: David Coulson <david [at] davidcoulson>
> > To: cisco-nsp [at] puck
> > Subject: [c-nsp] Stacking 3750X vs diverse 4948E
> > Message-ID: <4FB69B3D.3060802 [at] davidcoulson>
> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed
> >
> > In a datacenter environment, we typically deploy 4948 top-of-rack
> > switches with L2 uplinks to our 6500 core - Systems get connections into
> > two different switches and rely on OS NIC bonding (mostly Linux) to
> > support switch failures. Switches running STP and in the last four years
> > we've had no issues with this design (including failures of systems
> > connected to diverse switches).
> >
> > A new proposed configuration utilizes stacked 3750X switches, where
> > servers would be connected to multiple switches within the same stack. I
> > have next to no experience in the low-end switches that do stacking, but
> > from a general risk management perspective, it seems like a many eggs
> > and single basket configuration.
> >
> > Does anyone have any solid experience with 3750X switches, or stacking
> > in a datacenter in general? I've seen plenty of stacks for
> > closets/end-users, but I don't see many in a top-of-rack config. Is
> > Cisco stacking typically 'reliable', in that when a switch fails it will
> > leave the remainder of the stack functional? What about a software
> > issue? Does the whole stack crap out and reload, or does the master just
> > fail and a new one get elected?
> >
> > I realize it's a pretty broad question, but it boils down to - Is a
> > stacked switch config significantly less reliable/resilient/available
> > than two TOR switches?
> >
> > David
> >
> >
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


skeeve+cisconsp at eintellego

May 19, 2012, 9:57 PM

Post #19 of 40 (2785 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
If you're trying to eliminate STP, try Flexlink... which the Nexus 5010
doesn't support, but the 5500 series (apparently) does.
3750X does support Flexlink it seems.

*Skeeve Stevens, CEO*
eintellego Pty Ltd
skeeve [at] eintellego ; www.eintellego.net <http://www.eintellego.net.au>

Phone: 1300 753 383 ; Fax: (+612) 8572 9954

Cell +61 (0)414 753 383 ; skype://skeeve

facebook.com/eintellego

twitter.com/networkceoau ; www.linkedin.com/in/skeeve

PO Box 7726, Baulkham Hills, NSW 1755 Australia

The Experts Who The Experts Call
Juniper - Cisco – IBM



On Sun, May 20, 2012 at 1:26 PM, Tony Varriale <tvarriale [at] comcast>wrote:

> On 5/19/2012 6:47 AM, Lee wrote:
>
>> On 5/19/12, Saku Ytti<saku [at] ytti> wrote:
>>
>>> On (2012-05-18 14:55 -0400), David Coulson wrote:
>>>
>>> Does anyone have any solid experience with 3750X switches, or
>>>> stacking in a datacenter in general? I've seen plenty of stacks for
>>>>
>>> We've had quite many 3750 stacks, and we do see more problems in them
>>> than
>>> in other 3750 stacks.
>>> Particularly dangerous events are adding or removing members from stack.
>>>
>>> Also we've seen software defects which only affect stacks, like memory
>>> leakage on SNMP polling.
>>>
>>> Generally, do not add software complexity to the network if you have any
>>> practical alternative. STP while PITA, is tried and true and it actually
>>> does work in CSCO when configured right. So unless you absolutely need
>>> the
>>> extra interconnect capacity, go with STP.
>>>
>> How about VSS? We're considering it mainly because it would eliminate STP
>>
> VSS doesn't eliminate STP. Neither does vPC. So, you may want to
> reconsider.
>
> tv
> ______________________________**_________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/**mailman/listinfo/cisco-nsp<https://puck.nether.net/mailman/listinfo/cisco-nsp>
> archive at http://puck.nether.net/**pipermail/cisco-nsp/<http://puck.nether.net/pipermail/cisco-nsp/>
>
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


nsp.alexander.lim at gmail

May 19, 2012, 10:25 PM

Post #20 of 40 (2787 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
I think Nexus 5500 series support L3.

Regards,
Alexander Lim

On May 20, 2012, at 12:54 PM, Skeeve Stevens <skeeve+cisconsp [at] eintellego> wrote:

> Feature / Nexus 5010 / 3750X
>
> VLANs / 507 / 1005
> MAC / 16k / 4k-12k
> L3 / N / Y
> vPC / Y / N
>
> Nexus 5010 - less VLANs, no Layer 3, vPC
> 3750X - more VLAN, Layer 3, no vPC
>
>
> *Skeeve Stevens, CEO*
> eintellego Pty Ltd
> skeeve [at] eintellego ; www.eintellego.net <http://www.eintellego.net.au>
>
> Phone: 1300 753 383 ; Fax: (+612) 8572 9954
>
> Cell +61 (0)414 753 383 ; skype://skeeve
>
> facebook.com/eintellego
>
> twitter.com/networkceoau ; www.linkedin.com/in/skeeve
>
> PO Box 7726, Baulkham Hills, NSW 1755 Australia
>
> The Experts Who The Experts Call
> Juniper - Cisco – IBM
>
>
>
> On Sun, May 20, 2012 at 10:03 AM, scott owens <scottowens12 [at] gmail>wrote:
>
>> How about Nexus 5010s.
>>
>> I think they bundle them for less than 2 x 3750X .
>> We have both but the 3750s are used where we needed L2/L3, the 5Ks for just
>> L2 up to VSS or 7Ks.
>>
>>
>> you can boot them separately and they do LACP / Etherchannel just fine.
>>
>>
>>
>>
>>> 2. Stacking 3750X vs diverse 4948E (David Coulson)
>>> Message: 2
>>> Date: Fri, 18 May 2012 14:55:57 -0400
>>> From: David Coulson <david [at] davidcoulson>
>>> To: cisco-nsp [at] puck
>>> Subject: [c-nsp] Stacking 3750X vs diverse 4948E
>>> Message-ID: <4FB69B3D.3060802 [at] davidcoulson>
>>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>>
>>> In a datacenter environment, we typically deploy 4948 top-of-rack
>>> switches with L2 uplinks to our 6500 core - Systems get connections into
>>> two different switches and rely on OS NIC bonding (mostly Linux) to
>>> support switch failures. Switches running STP and in the last four years
>>> we've had no issues with this design (including failures of systems
>>> connected to diverse switches).
>>>
>>> A new proposed configuration utilizes stacked 3750X switches, where
>>> servers would be connected to multiple switches within the same stack. I
>>> have next to no experience in the low-end switches that do stacking, but
>>> from a general risk management perspective, it seems like a many eggs
>>> and single basket configuration.
>>>
>>> Does anyone have any solid experience with 3750X switches, or stacking
>>> in a datacenter in general? I've seen plenty of stacks for
>>> closets/end-users, but I don't see many in a top-of-rack config. Is
>>> Cisco stacking typically 'reliable', in that when a switch fails it will
>>> leave the remainder of the stack functional? What about a software
>>> issue? Does the whole stack crap out and reload, or does the master just
>>> fail and a new one get elected?
>>>
>>> I realize it's a pretty broad question, but it boils down to - Is a
>>> stacked switch config significantly less reliable/resilient/available
>>> than two TOR switches?
>>>
>>> David
>>>
>>>
>> _______________________________________________
>> cisco-nsp mailing list cisco-nsp [at] puck
>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/

_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


reuben-cisco-nsp at reub

May 19, 2012, 10:57 PM

Post #21 of 40 (2822 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
Yes - the 5548 does routing. We have 2x 5548UP's with the Layer 3
daughtercard in our small corporate DC.

It does routing, yes, but you need to be aware of caveats around the
feature. I suppose you could say that about any Cisco switch, but bear
in mind that NX-OS is aimed and targeted at the Data Centre market more
so than the Service Provider market so you you'll probably find it far
more feature rich in DC features such as Fibre Channel, than SP features
like EVCs (which btw it doesn't support).

You'll probably also find that there aren't a lot of choice of software
releases of code since the daughter card was introduced (and there are
now two variants of this, the second hardware revision has more onboard
resources). The quality of the code is reasonable but there aren't many
to choose from, and well, I'd be interested to know of anyone who has
really hammered the BGP code to the limits ;)

It's also nice to be able to go from 1G to 10G by just upgrading SFP's.

We've had to leave a 3560 in production to do IPv6 routing and policy
routing as it just can't be presently done on the Nexus 5000, and
there's no word when IPv6 routing will be supported on it, if ever.

Reuben



On 20/05/2012 3:25 PM, Alexander Lim wrote:
> I think Nexus 5500 series support L3.
>
> Regards,
> Alexander Lim
>
> On May 20, 2012, at 12:54 PM, Skeeve Stevens <skeeve+cisconsp [at] eintellego> wrote:
>
>> Feature / Nexus 5010 / 3750X
>>
>> VLANs / 507 / 1005
>> MAC / 16k / 4k-12k
>> L3 / N / Y
>> vPC / Y / N
>>
>> Nexus 5010 - less VLANs, no Layer 3, vPC
>> 3750X - more VLAN, Layer 3, no vPC
>>
>>
>> *Skeeve Stevens, CEO*
>> eintellego Pty Ltd
>> skeeve [at] eintellego ; www.eintellego.net <http://www.eintellego.net.au>
>>
>> Phone: 1300 753 383 ; Fax: (+612) 8572 9954
>>
>> Cell +61 (0)414 753 383 ; skype://skeeve
>>
>> facebook.com/eintellego
>>
>> twitter.com/networkceoau ; www.linkedin.com/in/skeeve
>>
>> PO Box 7726, Baulkham Hills, NSW 1755 Australia
>>
>> The Experts Who The Experts Call
>> Juniper - Cisco – IBM
>>
>>
>>
>> On Sun, May 20, 2012 at 10:03 AM, scott owens <scottowens12 [at] gmail>wrote:
>>
>>> How about Nexus 5010s.
>>>
>>> I think they bundle them for less than 2 x 3750X .
>>> We have both but the 3750s are used where we needed L2/L3, the 5Ks for just
>>> L2 up to VSS or 7Ks.
>>>
>>>
>>> you can boot them separately and they do LACP / Etherchannel just fine.
>>>
>>>
>>>
>>>
>>>> 2. Stacking 3750X vs diverse 4948E (David Coulson)
>>>> Message: 2
>>>> Date: Fri, 18 May 2012 14:55:57 -0400
>>>> From: David Coulson <david [at] davidcoulson>
>>>> To: cisco-nsp [at] puck
>>>> Subject: [c-nsp] Stacking 3750X vs diverse 4948E
>>>> Message-ID: <4FB69B3D.3060802 [at] davidcoulson>
>>>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>>>>
>>>> In a datacenter environment, we typically deploy 4948 top-of-rack
>>>> switches with L2 uplinks to our 6500 core - Systems get connections into
>>>> two different switches and rely on OS NIC bonding (mostly Linux) to
>>>> support switch failures. Switches running STP and in the last four years
>>>> we've had no issues with this design (including failures of systems
>>>> connected to diverse switches).
>>>>
>>>> A new proposed configuration utilizes stacked 3750X switches, where
>>>> servers would be connected to multiple switches within the same stack. I
>>>> have next to no experience in the low-end switches that do stacking, but
>>>> from a general risk management perspective, it seems like a many eggs
>>>> and single basket configuration.
>>>>
>>>> Does anyone have any solid experience with 3750X switches, or stacking
>>>> in a datacenter in general? I've seen plenty of stacks for
>>>> closets/end-users, but I don't see many in a top-of-rack config. Is
>>>> Cisco stacking typically 'reliable', in that when a switch fails it will
>>>> leave the remainder of the stack functional? What about a software
>>>> issue? Does the whole stack crap out and reload, or does the master just
>>>> fail and a new one get elected?
>>>>
>>>> I realize it's a pretty broad question, but it boils down to - Is a
>>>> stacked switch config significantly less reliable/resilient/available
>>>> than two TOR switches?
>>>>
>>>> David
>>>>
>>>>
>>> _______________________________________________
>>> cisco-nsp mailing list cisco-nsp [at] puck
>>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>>>
>> _______________________________________________
>> cisco-nsp mailing list cisco-nsp [at] puck
>> https://puck.nether.net/mailman/listinfo/cisco-nsp
>> archive at http://puck.nether.net/pipermail/cisco-nsp/
>
> _______________________________________________
> cisco-nsp mailing list cisco-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/
>


_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


saku at ytti

May 20, 2012, 1:36 AM

Post #22 of 40 (2782 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On (2012-05-19 22:25 -0500), Tony Varriale wrote:

> If you follow the rules, those are the easiest, most non-eventful
> events ever. I've done over 100 and had no issues.

This is curious statement, it implies that if you are operating devices as
per documentation, nothing ever goes wrong. But I'm sure this really wasn't
what you mean.
I can't explain why they've been uneventful for you, but eventful for us,
maybe since we've run it since 3750 shipped and are typically running very
old IOS (12.2(25)SE) (as virtually 0 problems on unstacked 3750, we've been
unmotivated to standardize on later releases).
Maybe your set of features configured in your 3750 in your IOS release is
better match. It would be arrogant to assume that non of our problems were
caused by operator mistakes, I'm sure some field-techs have done it wrong.

Now I'm not sure if our failure rate in stacking/destacking is 1% (1 in
100) it is probably less, all I'm saying they fail lot more often than our
non-stacked switches.

Quick bugtool search for 'stack 3750' gives 743 bugs. 569 if these have
been modified in last 6 months. 122 have been modified in last 3 months.

My key point here is, regardless how reliable or unreliable then new
software feature is, it will have more than 0 new problems, so unless you
actually can capitalize on that new feature, do not deploy it.
We found we don't need the extra capacity for inter-switch links, and we
did not get any real benefit from 'single' management plane. YMMV.


--
++ytti
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


david at davidcoulson

May 20, 2012, 4:49 AM

Post #23 of 40 (2751 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/19/12 11:24 PM, Tony Varriale wrote:
>
>
> The first and most important question is: Is this a real datacenter?
> If so, 3750xyz is not for you.
Yes, it is. Why not?
>
> Also, the regular 4948s are not much better.
>
We've been using 4948s for years. Never had any issues with them, beyond
the basic hardware problems. What sort of issues am I supposed to be
experiencing?

DAvid
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


peter at rathlev

May 20, 2012, 6:01 AM

Post #24 of 40 (2752 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On Sun, 2012-05-20 at 07:49 -0400, David Coulson wrote:
> On 5/19/12 11:24 PM, Tony Varriale wrote:
> > The first and most important question is: Is this a real datacenter?
> > If so, 3750xyz is not for you.
>
> Yes, it is. Why not?

I think Tony's question is about wether you actually need to move a lot
of data or just need to connect some servers. At work we have something
we often call "datacenters", but they're actually not. We don't move a
lot of data, have low density 10G and so forth. For us the 3560/3750
series has been adequate.

--
Peter


_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/


tvarriale at comcast

May 20, 2012, 7:35 PM

Post #25 of 40 (2720 views)
Permalink
Re: Stacking 3750X vs diverse 4948E [In reply to]
On 5/20/2012 3:36 AM, Saku Ytti wrote:
> On (2012-05-19 22:25 -0500), Tony Varriale wrote:
>
>> If you follow the rules, those are the easiest, most non-eventful
>> events ever. I've done over 100 and had no issues.
> This is curious statement, it implies that if you are operating devices as
> per documentation, nothing ever goes wrong. But I'm sure this really wasn't
> what you mean.

I typically operate networks as to what works. Any manufacturers
documentation is never 100% or what works best in the field. For
example, on 3750s I manually upgrade IOS to existing stack code rev.
> I can't explain why they've been uneventful for you, but eventful for us,

See above.
> Maybe your set of features configured in your 3750 in your IOS release is
> better match. It would be arrogant to assume that non of our problems were
> caused by operator mistakes, I'm sure some field-techs have done it wrong.

Arrogant of you? Or me? I wasn't implying anything. I was giving you
an opinion and my experience with 3750s.
>
> Now I'm not sure if our failure rate in stacking/destacking is 1% (1 in
> 100) it is probably less, all I'm saying they fail lot more often than our
> non-stacked switches.
Which part of the switches are failing?
>
> Quick bugtool search for 'stack 3750' gives 743 bugs. 569 if these have
> been modified in last 6 months. 122 have been modified in last 3 months.

Do you have NOS service? Or, do you/your team do bug scrubs? How about
any type of testing prior to going into production?

tv
_______________________________________________
cisco-nsp mailing list cisco-nsp [at] puck
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/

First page Previous page 1 2 Next page Last page  View All Cisco nsp RSS feed   Index | Next | Previous | View Threaded
 
 


Interested in having your list archived? Contact Gossamer Threads
 
  Web Applications & Managed Hosting Powered by Gossamer Threads Inc.