SG-1000 speed issues

[gama71]

Hi, i received my sg-1000 and have unfortunately the same result, troughput is really low.
i did a test with a celeron 1Ghz on a board with dual 100Mb ethernet , connexion was 60Mbits compare to direct connexion around 90. The sg-1000 is giving 18/ 35.

Too much loss, i was expecting a lot better :(.

i tries again, my connexion is a bit slow today,
SG-1000 : 8Mb down , 3 mb up
Celeron 1Ghz : 30Mb down, 3 mbup
direct connexion to modem  : 39 / 3

[stephenw10]

Are you seeing similar results to multiple test servers?

How about just downloading a file from a known high speed source?

I would not expect the SG-1000 to introduce any sort of a restriction on a connection that's less than 100Mbps.

Are you seeing any errors on the interfaces?

Any variation between snapshots?

Steve

[gama71]

Hi, i'm doing a simple speedtest with my internet provider (cable connection).
Speed is changing a bit from 30 to 40Mb , but i haave always the same ratio with the sg-1000, i'm around 33%.

i'm using version 2.3. on my celeron, the one on the sg-1000 is 2.4 beta

BR
Gaetan

[jimp]

Is it the same configuration on both systems? What features are enabled?

What is the CPU usage like on the SG-1000 during the test? ("top -aSH" from a shell prompt)

Are there any interface errors showing on Status > Interfaces? What does the interface speed/duplex show?

[skilbjo]

Hey there, I opened a ticket with customer support and we are working through some strategies. Something however that really helped was this:

Interfaces -> WAN -> General Configuration, change Speed and Duplex from automatic to 100base TX full-duplex. With that change, I've bumped up my speed test from 10-15mbs to 40-50mbps. I don't know why telling it explicitly which speed to use helped, but it did.

I am seeing CPU usage a little high during speed tests (~60% CPU usage), so we are going to debug a bit more to see if we can get it closer to the speed w/o the device (90mbs).

[matt1544]

Came here to report the same thing. My former "homebrew" pfsense box runs 130-140 Mbits/s without fail. However, as soon as I hooked up the SG-1000 (with same configuration otherwise) I drop to 15-30 Mbits/s. (See attached, before 1/7/17 vs after.)

This is probably actually fast enough for 85% of our normal residential use, but I would access to the full speeds I pay for.

I tried the WAN speed setting that skilbjo mentions above, but it didn't seem to do anything in quick testing. (but I'll try it for a few days and see what happens.)

Any other advice? Worth putting in a ticket for this?

thanks
Matt


[jimp]

See my post a few entries up and post the items I mentioned, also, post the output of this command:

sysctl -a | grep cpsw

[matt1544]

thanks. will get to the items you mentioned. meanwhile, here is the output of sysctl -a | grep cpsw from my box:

Shell Output - sysctl -a | grep cpsw
dev.etherswitch.0.%parent: cpswss0
dev.miibus.1.%parent: cpsw1
dev.miibus.0.%parent: cpsw0
dev.cpsw.1.%parent: cpswss0
dev.cpsw.1.%pnpinfo:
dev.cpsw.1.%location:
dev.cpsw.1.%driver: cpsw
dev.cpsw.1.%desc: Ethernet Switch Port
dev.cpsw.0.%parent: cpswss0
dev.cpsw.0.%pnpinfo:
dev.cpsw.0.%location:
dev.cpsw.0.%driver: cpsw
dev.cpsw.0.%desc: Ethernet Switch Port
dev.cpsw.%parent:
dev.cpswss.0.watchdog.resets: 0
dev.cpswss.0.queue.rx.longestChain: 0
dev.cpswss.0.queue.rx.queueRestart: 0
dev.cpswss.0.queue.rx.totalDequeued: 3838177
dev.cpswss.0.queue.rx.totalEnqueued: 3838561
dev.cpswss.0.queue.rx.maxAvailBuffers: 110
dev.cpswss.0.queue.rx.availBuffers: 0
dev.cpswss.0.queue.rx.maxActiveBuffers: 384
dev.cpswss.0.queue.rx.activeBuffers: 384
dev.cpswss.0.queue.rx.totalBuffers: 384
dev.cpswss.0.queue.tx.longestChain: 6
dev.cpswss.0.queue.tx.queueRestart: 0
dev.cpswss.0.queue.tx.totalDequeued: 4563228
dev.cpswss.0.queue.tx.totalEnqueued: 4563228
dev.cpswss.0.queue.tx.maxAvailBuffers: 128
dev.cpswss.0.queue.tx.availBuffers: 128
dev.cpswss.0.queue.tx.maxActiveBuffers: 8
dev.cpswss.0.queue.tx.activeBuffers: 0
dev.cpswss.0.queue.tx.totalBuffers: 128
dev.cpswss.0.stats.GoodRxFrames: 3880465
dev.cpswss.0.stats.BroadcastRxFrames: 14793
dev.cpswss.0.stats.MulticastRxFrames: 40445
dev.cpswss.0.stats.PauseRxFrames: 0
dev.cpswss.0.stats.RxCrcErrors: 293
dev.cpswss.0.stats.RxAlignErrors: 103
dev.cpswss.0.stats.OversizeRxFrames: 0
dev.cpswss.0.stats.RxJabbers: 0
dev.cpswss.0.stats.ShortRxFrames: 0
dev.cpswss.0.stats.RxFragments: 4479
dev.cpswss.0.stats.RxOctets: 3514843195
dev.cpswss.0.stats.GoodTxFrames: 4009902
dev.cpswss.0.stats.BroadcastTxFrames: 236
dev.cpswss.0.stats.MulticastTxFrames: 1960
dev.cpswss.0.stats.PauseTxFrames: 0
dev.cpswss.0.stats.DeferredTxFrames: 0
dev.cpswss.0.stats.CollisionsTxFrames: 0
dev.cpswss.0.stats.SingleCollisionTxFrames: 0
dev.cpswss.0.stats.MultipleCollisionTxFrames: 0
dev.cpswss.0.stats.ExcessiveCollisions: 0
dev.cpswss.0.stats.LateCollisions: 0
dev.cpswss.0.stats.TxUnderrun: 0
dev.cpswss.0.stats.CarrierSenseErrors: 0
dev.cpswss.0.stats.TxOctets: 3559792735
dev.cpswss.0.stats.RxTx64OctetFrames: 69663
dev.cpswss.0.stats.RxTx65to127OctetFrames: 2387733
dev.cpswss.0.stats.RxTx128to255OctetFrames: 320328
dev.cpswss.0.stats.RxTx256to511OctetFrames: 344107
dev.cpswss.0.stats.RxTx512to1024OctetFrames: 334674
dev.cpswss.0.stats.RxTx1024upOctetFrames: 4434261
dev.cpswss.0.stats.NetOctets: 7074772555
dev.cpswss.0.stats.RxStartOfFrameOverruns: 2
dev.cpswss.0.stats.RxMiddleOfFrameOverruns: 0
dev.cpswss.0.stats.RxDmaOverruns: 0
dev.cpswss.0.ports.1.uptime: 14359
dev.cpswss.0.ports.0.uptime: 14361
dev.cpswss.0.intr_coalesce_us: 0
dev.cpswss.0.attachedSecs: 14505
dev.cpswss.0.rx_batch: 0
dev.cpswss.0.debug: 0
dev.cpswss.0.%parent: simplebus0
dev.cpswss.0.%pnpinfo: name=ethernet@4a100000 compat=ti,am335x-cpsw
dev.cpswss.0.%location:
dev.cpswss.0.%driver: cpswss
dev.cpswss.0.%desc: 3-port Switch Ethernet Subsystem
dev.cpswss.%parent:

[jimp]

dev.cpswss.0.stats.RxCrcErrors: 293
dev.cpswss.0.stats.RxAlignErrors: 103

That would seem to imply that one of the ports (unfortunately it doesn't know which) was receiving some bad data from its peer. Though in other "slow" cases we've seen these counters much, much higher.

It could be an indication of a cabling or speed negotiation problem, but it's usually at L1 or L2. Swap cables/ports, check speed/duplex, etc.

[W4RH34D]

These things are perfect for our redundant T1 lines.

Check sig.