Switch packet Buffer size

[Moudar]

Hi,

Two switch models, 9200l and 1000 series switches.

9200l buffer size:

6 MB buffers for 24- or 48- port Gigabit Ethernet models, 12 MB buffers for 24 or 48 port multigigabit models.

1000 switch buffer size:

The egress buffer is 1.5 MB on all Cisco Catalyst 1000 Series Switches.

The question is how can this difference in buffer size affect the network. If you can explain in short what do these buffers do is best!

Our network has very different places and different workload, some buildings may have 10 48port switches (about 200 users) and other places can have 8 port switch with one user, one AP and a printer.

The next question, we are planning to invest in 1000 switches because they are cheaper and we need many (about 400) we don't plan to invest in DNA center because it is expensive. Maybe we will buy DNA when Prime vanishes and then we will use the light version with no switch support only to save money.

The question is, what is the difference between 9200l and 1000 switches if we don't count DNA center support and price?

We have some 9200l in production today and what i can see is some interface drops as example, Total output drops

GigabitEthernet1/0/6 is up, line protocol is up (connected)
  Hardware is Gigabit Ethernet, address is 8c84.4248.6806 (bia 8c84.4248.6806)
  MTU 1500 bytes, BW 1000000 Kbit/sec, DLY 10 usec,
     reliability 255/255, txload 1/255, rxload 1/255
  Encapsulation ARPA, loopback not set
  Keepalive set (10 sec)
  Full-duplex, 1000Mb/s, media type is 10/100/1000BaseTX
  input flow-control is on, output flow-control is unsupported
  ARP type: ARPA, ARP Timeout 04:00:00
  Last input 00:00:22, output 00:00:00, output hang never
  Last clearing of "show interface" counters 3w0d
  Input queue: 0/2000/0/0 (size/max/drops/flushes); Total output drops: 1592
  Queueing strategy: fifo
  Output queue: 0/40 (size/max)
  5 minute input rate 79000 bits/sec, 20 packets/sec
  5 minute output rate 78000 bits/sec, 29 packets/sec
     29146158 packets input, 10155311111 bytes, 0 no buffer
     Received 19175 broadcasts (17955 multicasts)
     0 runts, 0 giants, 0 throttles
     0 input errors, 0 CRC, 0 frame, 0 overrun, 0 ignored
     0 watchdog, 17955 multicast, 0 pause input
     0 input packets with dribble condition detected
     37372686 packets output, 19854211284 bytes, 0 underruns
     Output 171449 broadcasts (623559 multicasts)
     0 output errors, 0 collisions, 0 interface resets
     237 unknown protocol drops
     0 babbles, 0 late collision, 0 deferred
     0 lost carrier, 0 no carrier, 0 pause output
     0 output buffer failures, 0 output buffers swapped out

 Do these drops have anything with the switch buffer size?

[balaji.bandi]

The question is how can this difference in buffer size affect the network. If you can explain in short what do these buffers do is best!

Switches have large frame buffers that allow data waiting to be transmitted to be stored so the data will not be dropped. This feature is beneficial especially if the incoming traffic is from a faster port than the egress port used for transmitting.

what is the difference between 9200l and 1000 switches  <-- both are different model, Once used for enterprise Grade, another small business and Enterprice in the middle. (you can find the difference in Datasheet ) - about each model and their backplate process.

We have some 9200l in production today and what i can see is some interface drops as an example, Total output drops   < -- this can be any reason

First suggestion clean the counters an d check how fast they grow ? this also indirectly related  utilisation of interface.

https://www.cisco.com/c/en/us/support/docs/switches/catalyst-9300-switch/216236-troubleshoot-output-drops-on-catalyst-90.html

[Moudar]

9200l switch ports are connected only to laptops which have 1G ethernet port. The drops are very small as shown in the above picture, but still there are drops and need to understand why they are happen. I have already reset the counters and still have drops on many ports not few

[Joseph W. Doherty]

"Do these drops have anything with the switch buffer size?"

Possibly not, because:

0 output buffer failures

more likely due to:

"The question is how can this difference in buffer size affect the network."

If can cause egress queuing drops from lack of buffers.

@balaji.bandi describes: "This feature is beneficial especially if the incoming traffic is from a faster port than the egress port used for transmitting.", which is true, but the other cause is multiple ingress interfaces sending to the same egress interface.  If multiple interfaces' aggregate bandwidth exceeds the egress interface's bandwidth (just as with a single higher bandwidth ingress interface), buffering/queuing is often needed.

[Moudar]

So, is the chance to have drops on 1000 switch port bigger than to have it on 9200l switch port?

As I said we have already some drops on the 9200l switches, if we choose to use 1000 switches, will these drops increase ?

In other words, if i change my 9200l switch that have some drops with other new 1000 switch, (both are 24-ports with 1G uplink) Would the drops increase because of the smaller buffer of the 1000 series switch?

[Joseph W. Doherty]

To your three questions, all a variation of, will drops increase moving from 9200L to 1000 switches, cannot easily say.

Why?

Because, drop rates are not only impacted by physical buffer resources, by also by how the switch's architecture uses its buffers and how the switch is configured to use its buffers.

An example of the architecture.  Some switches set up their buffers as a global pool that all ports use as needed.  Some switches allocate a fix amount of buffers per port.  Some switches have both a common buffer pool and buffer reservations per port, and for the latter, the switch might support configuration changes on allocations percentages, between the common buffer pool and per port allocations, or even specific allocations per port.

An example of configuration impact.  Switches often have "logical" resource limits imposed before they actually use all the buffer resources.  These logical limits might be modified by configuration and might have a lot of options that go hand-in-hand with QoS features.  For example, when you (QoS) prioritize one class of traffic over another, there's two schools of thought.  One is, the prioritized traffic class should have larger logical limits to help insure that class's traffic is not dropped.  The other is, as the traffic is prioritized for transmission, it shouldn't need larger buffer limits.

The forgoing tries to show why I cannot easily answer your question without fully understanding the architectures of the two switch series (which I do not), without knowing all the possible configuration options (which I also do not know) and also how your switches are currently configured and/or whether such configuration is actually optimal for your traffic on that switch (which I also do not know).

Keeping all the above in mind, a "better" grade of switch, often has more resources and/or configuration options so that optimally configured it can perform better that a "lower" grade of switch.  Yet, YOUR traffic might not need the "better" grade of switch.

[MHM Cisco World]

each UADP that use to forward traffic between ingress and egress port have Buffer size, this size determine the Total BW of traffic INSIDE the SW.
so Yes the buffer is effect the SW performance