SG-300 slow speeds

[has123456]

Hi there,

I've just purchased two SG300-28 switches. The setup is as follows:

SG300-1 (IP 192.168.1.30 VLAN1) Connects to HP 10/100 Core switch via GE26 . 

SG300-2 (IP 192.168.1.30VLAN1) Connects to HP 10/100 Core switch via GE26

The data and management network are both VLAN1, 192.168.1.0/24 subnet.

My server (2012R2) has windows nic teaming enabled with one nic in switch1 and the other in switch2. Both ports are VLAN1 and the nic team has an IP of 192.168.1.56. Server nics are 1 GBe.

While the switches have other VLANs on 2 ports, they are local to the switch only and are not trunked across on GE26.

The default gateway is 192.168.1.254. I do not have access to the HP Core switch.

All ports on the SG's are in trunk mode.

The issue I am having is that network speeds are very slow, averaging from 10 to 20 MB/S during large file transfers.

I have turned off QOS. I have enabled and disabled Jumbo frames and I have turned STP off and then on as well with no difference. I have broken the nic team and tried to transfer data to the server with my laptop connected directly to one of the switches but I'm not seeing any real improvement (when laptop is connected to switch one and server is on switch2, I see better performance than when I have them on the same switch).

Now if I instead disconnect SG300-1 from the HP and connect it to the SG300-2 and connect that to the HP Core switch, my speeds jump up to 600 to 800 MB/s all across the board.

Can anyone please advise further as I'm a bit stumped and am not that experienced when it comes to networking?

Thanks

HA

[rakeshvelagala]

Hi,

SG300-1 (IP 192.168.1.30 VLAN1) Connects to HP 10/100 Core switch via GE26 . 

SG300-2 (IP 192.168.1.30VLAN1) Connects to HP 10/100 Core switch via GE26

Both have same IP?Are they in Stack?

[rakeshvelagala]

I believe SG-300 you cannot stack.

Please advise if you have done as below.

http://blogs.technet.com/b/privatecloud/archive/2012/06/19/nic-teaming-in-windows-server-2012-brings-simple-affordable-traffic-reliability-and-load-balancing-to-your-cloud-workloads.aspx

Thanks

[has123456]

Hi,

Thanks for replying. Sorry I made a typo.. The switches are not stacked. The SG300-1 is 192.168.1.30 and the SG300-2 is 192.168.1.31.

The link you have referenced is the one I have followed. The diagram depicting my setup is the 'Host with interfaces connected to different switches' one.

Thanks,

HA

[rakeshvelagala]

Hi,

Can you please take a backup of the switches and post the config?

Thanks

[rakeshvelagala]

Hi 

Disable teaming. Connect the server to only SW2.

Since they are on the same subnet, can you please check if the server Mac address is reflected correctly on the host PC when connected to switch1?

arp -a

Can you do a continuous ping to your server and advise if there are any drops from the host in Switch1 to Server.

I believe your setup is as below.

Host------SW1------core switch------Sw2------Server 

Since they are in the same subnet, they should be L2 switched.

Are the Ports connected to Host in access mode?

Thanks

[has123456]

Hi,

Thanks for your reply. Please find the config details for both switches:

config-file-header

sg300-1

v1.3.7.18 / R750_NIK_1_35_647_358

CLI v1.0

set system mode switch

file SSD indicator excluded

@

port jumbo-frame

vlan database

vlan 60-61

exit

voice vlan oui-table add 0001e3 Siemens_AG_phone________

voice vlan oui-table add 00036b Cisco_phone_____________

voice vlan oui-table add 00096e Avaya___________________

voice vlan oui-table add 000fe2 H3C_Aolynk______________

voice vlan oui-table add 0060b9 Philips_and_NEC_AG_phone

voice vlan oui-table add 00d01e Pingtel_phone___________

voice vlan oui-table add 00e075 Polycom/Veritel_phone___

voice vlan oui-table add 00e0bb 3Com_phone______________

no eee enable

no qos

hostname sg300-1

username cisco password encrypted <removed by me> privilege 15

ip ssh server

snmp-server location "Server Room"

clock timezone " " 0 minutes 0

clock source sntp

sntp unicast client enable

sntp unicast client poll

sntp server time-a.timefreq.bldrdoc.gov poll

sntp server time-b.timefreq.bldrdoc.gov poll

ip telnet server

!

interface vlan 1

 ip address 192.168.1.30 255.255.255.0

!

interface vlan 60

 name ClusterTraffic

!

interface vlan 61

 name LiveMigration

!

interface gigabitethernet6

 switchport trunk native vlan 60

!

interface gigabitethernet7

 switchport trunk native vlan 61

!

interface gigabitethernet18

 switchport trunk native vlan 60

!

interface gigabitethernet19

 switchport trunk native vlan 61

!

interface Port-channel1

 description LAGtoCoreSwitch

!

exit

banner login 

SG300-1



macro auto disabled

-----------------

config-file-header

sg300-2

v1.3.7.18 / R750_NIK_1_35_647_358

CLI v1.0

set system mode switch

file SSD indicator excluded

@

port jumbo-frame

vlan database

vlan 60-61

exit

voice vlan oui-table add 0001e3 Siemens_AG_phone________

voice vlan oui-table add 00036b Cisco_phone_____________

voice vlan oui-table add 00096e Avaya___________________

voice vlan oui-table add 000fe2 H3C_Aolynk______________

voice vlan oui-table add 0060b9 Philips_and_NEC_AG_phone

voice vlan oui-table add 00d01e Pingtel_phone___________

voice vlan oui-table add 00e075 Polycom/Veritel_phone___

voice vlan oui-table add 00e0bb 3Com_phone______________

no eee enable

no boot host auto-config

no qos

hostname sg300-2

username cisco password encrypted <Removed by me> privilege 15

ip ssh server

snmp-server location "Server Room"

clock timezone " " 0 minutes 0

clock source sntp

sntp unicast client enable

sntp unicast client poll

sntp server time-a.timefreq.bldrdoc.gov poll

sntp server time-b.timefreq.bldrdoc.gov poll

sntp server time.windows.com poll

ip telnet server

!

interface vlan 1

 ip address 192.168.1.31 255.255.255.0

!

interface vlan 60

 name ClusterTraffic

!

interface vlan 61

 name LiveMigration

!

interface gigabitethernet6

 switchport trunk native vlan 60

!

interface gigabitethernet7

 switchport trunk native vlan 61

!

interface gigabitethernet18

 switchport trunk native vlan 60

!

interface gigabitethernet19

 switchport trunk native vlan 61

!

interface Port-channel1

 description LAGtoCoreSwitch

!

exit

banner login 

SG300-2



macro auto disabled 

Currently, the setup is as follows:

Switch1--Core Switch-- Switch2 

There are a couple of other servers connected to the switches as well with similalr issues so I haven't included them in my example above. When I plug a laptop into switch 2, with nic teaming this becomes:

Server1 nic1 ---Switch1---Core Switch---Switch2---Server1 Nic2 & laptop

I'll be on site tomorrow morning so will try to disable the nic teaming on Server1 and test the following:

Laptop ---- Switch1 ---Core Switch ---Switch2 --- Server1 nic1.

Is this the config you would like  me to test?

All ports on the switches are currently in trunk mode (default).

The LAG reference that you see in the config is the LAG I initially created, but then I removed the members (GE26) from the LAG as I did not have access to the Core switch. If the Core switch is running in L2 mode as well, could the switching be happening there (and resulting in a lower speed as it is 10/100)?

 Thanks,

HA

[has123456]

config-file-header

sg300-1

v1.3.7.18 / R750_NIK_1_35_647_358

CLI v1.0

set system mode switch

file SSD indicator excluded

@

port jumbo-frame

vlan database

vlan 60-61

exit

voice vlan oui-table add 0001e3 Siemens_AG_phone________

voice vlan oui-table add 00036b Cisco_phone_____________

voice vlan oui-table add 00096e Avaya___________________

voice vlan oui-table add 000fe2 H3C_Aolynk______________

voice vlan oui-table add 0060b9 Philips_and_NEC_AG_phone

voice vlan oui-table add 00d01e Pingtel_phone___________

voice vlan oui-table add 00e075 Polycom/Veritel_phone___

voice vlan oui-table add 00e0bb 3Com_phone______________

no eee enable

no qos

hostname sg300-1

username cisco password encrypted <removed by me> privilege 15

ip ssh server

snmp-server location "Server Room"

clock timezone " " 0 minutes 0

clock source sntp

sntp unicast client enable

sntp unicast client poll

sntp server time-a.timefreq.bldrdoc.gov poll

sntp server time-b.timefreq.bldrdoc.gov poll

ip telnet server

!

interface vlan 1

 ip address 192.168.1.30 255.255.255.0

!

interface vlan 60

 name ClusterTraffic

!

interface vlan 61

 name LiveMigration

!

interface gigabitethernet6

 switchport trunk native vlan 60

!

interface gigabitethernet7

 switchport trunk native vlan 61

!

interface gigabitethernet18

 switchport trunk native vlan 60

!

interface gigabitethernet19

 switchport trunk native vlan 61

!

interface Port-channel1

 description LAGtoCoreSwitch

!

exit

banner login 

SG300-1



macro auto disabled

-----------------

config-file-header

sg300-2

v1.3.7.18 / R750_NIK_1_35_647_358

CLI v1.0

set system mode switch

file SSD indicator excluded

@

port jumbo-frame

vlan database

vlan 60-61

exit

voice vlan oui-table add 0001e3 Siemens_AG_phone________

voice vlan oui-table add 00036b Cisco_phone_____________

voice vlan oui-table add 00096e Avaya___________________

voice vlan oui-table add 000fe2 H3C_Aolynk______________

voice vlan oui-table add 0060b9 Philips_and_NEC_AG_phone

voice vlan oui-table add 00d01e Pingtel_phone___________

voice vlan oui-table add 00e075 Polycom/Veritel_phone___

voice vlan oui-table add 00e0bb 3Com_phone______________

no eee enable

no boot host auto-config

no qos

hostname sg300-2

username cisco password encrypted <Removed by me> privilege 15

ip ssh server

snmp-server location "Server Room"

clock timezone " " 0 minutes 0

clock source sntp

sntp unicast client enable

sntp unicast client poll

sntp server time-a.timefreq.bldrdoc.gov poll

sntp server time-b.timefreq.bldrdoc.gov poll

sntp server time.windows.com poll

ip telnet server

!

interface vlan 1

 ip address 192.168.1.31 255.255.255.0

!

interface vlan 60

 name ClusterTraffic

!

interface vlan 61

 name LiveMigration

!

interface gigabitethernet6

 switchport trunk native vlan 60

!

interface gigabitethernet7

 switchport trunk native vlan 61

!

interface gigabitethernet18

 switchport trunk native vlan 60

!

interface gigabitethernet19

 switchport trunk native vlan 61

!

interface Port-channel1

 description LAGtoCoreSwitch

!

exit

banner login 

SG300-2



macro auto disabled 

[has123456]

Hi,

All the ports on both switches are in trunk mode (default setting).

If the core switch is an L2 Switch, would the switching happen there? It is 10/100 and I'm wondering if that would affect the speed.

The current setup is as follows (with teaming):

Server nic1--Switch1--core switch--Switch2--Server nic2

With a laptop in switch 2, it becomes:

Server nic1--Switch1--core switch--Switch2--Server nic2 & laptop 

Am I correct in that you want me to disable nic teaming on the server and try this:

Laptop nic---Switch1---Core Switch--Switch2---Server nic1 ?

Thanks,

HA

[Charles Hill]

I have seen speed duplex mismatch can cause a slowness in the network.

Verify both sides are either hard coded or both sides are set to auto on the switch ports and server ports.

Another issue with slowness in downloading large files from a server can be caused by the large send offload feature on the nic.

Try disabling large send offload on the nic properties.

http://www.peerwisdom.org/2013/04/25/disabling-large-send-offload-windows/

[has123456]

Hi,

I set the ports to access mode (except the trunk port -GE26). I broke teaming and saw no difference in speed (10-11 MB/s).

I then plugged in a spare SG300-10 so that:

Laptop--Switch1-(SG300-10)--Core Switch--(SG300-10)--Switch2--Server

With the SG300-10 in between the core switch and the 2 SG300-28 switches, my speed jumped up to 950 MB/s.

I then turned on teaming and was able to get identical speeds. When I transferred 2 large files between 2 servers  (both with teaming), I got 950MB/s upload and around 930 MB/s download at the same time.

I then disconnected the SG300-10 and connected both SG300-28's to the core as before.

Transfers were capped at 20 MB/s

I also managed to get access to the core switch. The ports on that are 10/100. The two SG300-28's are connected to ports 18 and 19 on the core switch. When I transferred data between two servers with teamed nics, I saw that traffic on ports 18 and 19 on the core switch reached 100% which would equate to around 20MB/S combined which is what I am seeing when transferring data. So it seems that the data from both servers is passing through the core switch and slowing down.

I would have thought that switch independent teaming in this case would mean that as nic1 on server 1 and nic1 on server 2 are connected to the same switch (SG300-1), data would not traverse to the Core Switch but instead stay on the SG300-1. Only if the nic1 on on any of the servers go down, the data would traverse the core switch.

Thanks,

HA

[rakeshvelagala]

Hi 

Can you please paste the output of   #show interface  for those ports(18 and 19) from the core switch when SG300-28 is connected?

It looks like a duplex mismatch as  Cehill pointed out.

Frankly speaking connecting SG300-10 should not make a difference.

Thanks

[has123456]

Hi,

Thanks for your email. Please find the output below:

HP ProCurve Switch 2626# show interface 18

 Status and Counters - Port Counters for port 18

  Name  :

  Link Status     : Up

  Bytes Rx        : 1,204,203,031       Bytes Tx        : 2,560,540,352
  Unicast Rx      : 31,817,698          Unicast Tx      : 44,684,962
  Bcast/Mcast Rx  : 2,868,610           Bcast/Mcast Tx  : 9,222,398

  FCS Rx          : 0                   Drops Rx        : 45,725
  Alignment Rx    : 0                   Collisions Tx   : 0
  Runts Rx        : 0                   Late Colln Tx   : 0
  Giants Rx       : 0                   Excessive Colln : 0
  Total Rx Errors : 0                   Deferred Tx     : 0

HP ProCurve Switch 2626# show interface 19

 Status and Counters - Port Counters for port 19

  Name  :

  Link Status     : Up

  Bytes Rx        : 2,297,691,560       Bytes Tx        : 1,786,431,836
  Unicast Rx      : 40,023,330          Unicast Tx      : 31,665,385
  Bcast/Mcast Rx  : 3,302,125           Bcast/Mcast Tx  : 8,797,827

  FCS Rx          : 1                   Drops Rx        : 24,350
  Alignment Rx    : 3                   Collisions Tx   : 0
  Runts Rx        : 0                   Late Colln Tx   : 0
  Giants Rx       : 0                   Excessive Colln : 0
  Total Rx Errors : 4                   Deferred Tx     : 0

When I have only one of the SG300's running, data speeds are fine. It's when both are running that the speed issue occurs and data seems to go through the core switch (which is layer 2).

Thanks,

HA

[has123456]

Hi rakeshvelagala & Cehill,

Thanks for your help on this.

Just posting this in case someone comes across a similar issue. This seems to be the result of Windows 2012R2 native nic teaming. In switch independent mode, the teaming logic is handled by Windows. To be more specific, when using address hashing as the teaming method, we cannot determine for example, that nic1 on server 1 will talk to nic 1 on server 2 (both on same switch). nic1 on server 1 (and switch1) may talk to nic2 on server 2 (switch2) and therefore go through the core switch which is 10/100. To verify, I put nic 2 on both servers to standby mode instead of active and saw speeds jump up to 950MB/s. Same results when putting nic1 on both hosts as standby and making nic2 active. Further details here: http://www.aidanfinn.com/?p=14032

Regards,

HA

[Charles Hill]

Thank you for the update.