Quanta LB6M (10GbE) -- Discussion

dwright1542

Active Member
Dec 26, 2015
371
73
28
49
I see from elsewhere in the thread that I paid too much for my lb6m... Oh Well...
Meh. I paid $350 a few months back. Still a steal for a really decent high speed switch. I've been buying them a putting them in redundant pairs. They do exactly what I need them to do.
 

Caleb

Member
Nov 16, 2015
39
8
8
34
Meh. I paid $350 a few months back. Still a steal for a really decent high speed switch. I've been buying them a putting them in redundant pairs. They do exactly what I need them to do.
Buying them in pairs? How many do you need? Would love to hear about your infrastructure. Also, where are you housing these suckers as they are loud too!
 

dwright1542

Active Member
Dec 26, 2015
371
73
28
49
Buying them in pairs? How many do you need? Would love to hear about your infrastructure. Also, where are you housing these suckers as they are loud too!
We've been installing them in LOTS of installs, many hyperconverged VMware environments. So backend iSCSI / VMotion mostly. Basically a self-insured switch policy.
 
  • Like
Reactions: Chuntzu

keoki

New Member
Jun 2, 2016
21
19
3
60
We've been installing them in LOTS of installs, many hyperconverged VMware environments. So backend iSCSI / VMotion mostly. Basically a self-insured switch policy.
So what kind of performance do you see? Have you done any Bonnie++ benchmarks?

I have a FreeNAS box that I built a year ago, and it has a bunch of NFS mounts for vmware and daily software build storage. I recently added a pair of 10g ports feeding a couple of Cisco 6509 switches, pushing iSCSI to VMware systems. In a benchmark I ran the other day, I was seeing 715MB/s writes and 299MB/s reads, and iops averaging around 12,000. The existing loads on the SAN are obviously light, and the filesystem is only 1/4th full, so the benchmark system is the only stress point, and was mainly just testing the ram based ARC, so the benchmark wasn't testing the filesystem out to the spindle. I plan to do some more extensive benchmarking once I have 4 of my VMware servers on 10g iSCSI... I figure I can loop the benchmarks on 2 vm's at a time on each server, and with 8 sets of benchmarks running, as some scripts also fill the disk with random size files, I should be able to get an idea of how the unit will hold up if I let it fill up with data while under a load. I can also tune the ARC down in size to force the I/O load all the way to the spindles, but in MY real world there will never be that much load.

Really, I just want to be able to have a burst of activity every once in a while without saturating the NAS system... and that is pretty much what the benchmark tests for.

I'm told I made a critical error using RaidZ3 instead of striped mirrors, but the design goal is modest performance and maximum reliability, so the FreeNAS guys think I'm being stupid...

Fynny thing, when people act like I'm stupid, and say what I'm talking about is impossible... Like in 1990 I was told it was /impossible/ to sell ISP services to the general public because of the amount of support required... And the person that told me that knew I had been doing it for 3 years by then. And he was a published internet expert. I love it when people tell me to my face that the thing I am successfully demonstrating is impossible.

Perhaps I *should* scream like Godzilla...
 
  • Like
Reactions: Sleyk

keoki

New Member
Jun 2, 2016
21
19
3
60
Has anyone had any luck with 1G copper SFPs in the SFP+ slots? I have tried Finisar FCLF-8521-3, eNet GLC-T-ENC, and Mikrotik S-RJ01, but none of them are working.
I tried this on one of my Cisco 6509's, using Cisco SFP's, and they did *nothing*. I honestly don't think that is a valid thing to do. I was trying this because I needed to feed 1G fiber to a GPON router. They originally asked for 10G, so that is what I was expecting, and when the 10G wasn't working, I looked at his SFP, and saw it was only a 1g SFP... so we tried every model of 1g sfp in the 6509 10g SFP+ ports, and failed to get anything to work.

In my work lab, I don't have a lot of SFP ports, except on my impairment generator and DSLAMs, but on both of those I'm using copper SFP's as I don't have a valid application for 1G fiber ethernet, except where I feed ethernet cables into my RF chamber.... Most of my switches only have SFP+. Obviously I use fiber for most of the 10G connections, but I do have a few twinnax copper 10g cables for things physically close to the switches.
 

Sleyk

Your Friendly Knowledgable Helper and Techlover!
Mar 25, 2016
1,361
707
113
Stamford, CT
Keoki, keep doing the impossible! Oh, and Welcome to STH! We are glad to have ya!
 

pgh5278

Active Member
Oct 25, 2012
481
130
43
Australia
I paid over $100 more than others.....keoki, lets both scream like Godzilla.....
But you probably used it longer...doesnt matter what others paid, if you felt like you got a good deal, be happy.. Some people pay 3 or 4 times more to fly in the next seat on an airplane, as they booked it to early or too late for that flight, other flights it is good to buy late if there a lots of empty seats. Life is full of these experiences. You could have waited and perhaps caused more frustration by missing out..LOL..
 

Terry Kennedy

Well-Known Member
Jun 25, 2015
1,123
573
113
New York City
www.glaver.org
I suspect we will need dual rate SFP+ modules to even attempt 1G.
A while back in this thread, I posted a rather long explanation about how speed conversion won't happen in SFP/SFP+ modules (other than in some GbE TX modules). It is probably worth a re-read.

The dual speed 1/10 optics simply report that they support both speeds. The hardware in the switch also needs to support both speeds. And even if that is the case, there's no guarantee that dual-speed optics will work - for example, I have an Dell 8024 switch that supports both SFP and SFP+, but it will not work with a dual-speed optic part, even a genuine Dell one for a different model of switch.
 

Sleyk

Your Friendly Knowledgable Helper and Techlover!
Mar 25, 2016
1,361
707
113
Stamford, CT
But you probably used it longer...doesnt matter what others paid, if you felt like you got a good deal, be happy.. Some people pay 3 or 4 times more to fly in the next seat on an airplane, as they booked it to early or too late for that flight, other flights it is good to buy late if there a lots of empty seats. Life is full of these experiences. You could have waited and perhaps caused more frustration by missing out..LOL..
Nah man, for me, I gotta feel like I paid a reasonable price for things. I'm a cheap, clown bastard. I have only had my switch since February, and now people are getting these bad boys for almost half of what i paid for it a few months later. Makes me wanna blow fire like Godzilla too.

That feeling sucks monkey balls dipped in feces mixed with dried up fruit and bits of dead leaves man!

In retrospect, I didn't feel like I got a good deal to begin with, but I wanted 10Gb so badly, I forked over the dead Commanders in Chiefs. Ah well, such is life. I don't think I will ever be happy knowing I could have gotten the switch much cheaper though. I must say, it works, so that's all that matters.

Well, mostly all that matters. I could've bought a few more 2TB drives with that saved money man!

Shit on a stick with rice and beans, mixed into stew and mistaken for chili!
 

Chris Buchanan

New Member
Jun 6, 2016
5
0
1
45
Hello all! I recently just got my quanta lb6m switch in (like, an hour ago), and it works great so far. I can not login to the switch however with the admin and blank password combo. Is there a way to reset the switch? I can telnet to it and access a web interface.
 

Toby

New Member
May 6, 2016
17
4
3
48
Hello all! I recently just got my quanta lb6m switch in (like, an hour ago), and it works great so far. I can not login to the switch however with the admin and blank password combo. Is there a way to reset the switch? I can telnet to it and access a web interface.
Are you sure you have an LB6M? The LB6M doesn't have a web interface.
 

Chris Buchanan

New Member
Jun 6, 2016
5
0
1
45
Well crap. Turns out I was telnetting to the wrong IP. One should not buy multiple switches and put them on the network at the same time. That is my bad. I managed to get in.

(FASTPATH Routing) #show version

Switch: 1

System Description............................. Quanta LB6M, 1.2.0.14, Linux
2.6.21.7

(FASTPATH Routing) #show bootvar

Image Descriptions

active :
backup :


Images currently available on Flash

--------------------------------------------------------------------
unit active backup current-active next-active
--------------------------------------------------------------------

1 1.2.0.14 1.0.0.17 1.2.0.14 1.2.0.14

How does LACP work for everyone? Is it worth it? And Ive read on this thread that jumbo frames are hit or miss. Is it really worth it to have jumbo frames with 10gb?
 

Terry Kennedy

Well-Known Member
Jun 25, 2015
1,123
573
113
New York City
www.glaver.org
How does LACP work for everyone? Is it worth it?
It depends on your use case. A single session (flow of packets) is still going to be limited to the speed of a port. You need to have multiple sessions in order to get any benefit from LACP. And you need to select a balancing algorithm (probably called hashing-mode on the LB6M) that does the best job of distributing your sessions among the ports that make up the LACP bundle.

LACP is normally used between switches and in other backbone-type applications, rather than between end nodes and the switch, for the above reasons.
And Ive read on this thread that jumbo frames are hit or miss. Is it really worth it to have jumbo frames with 10gb?
This is another "it depends". Jumbo frames were designed to reduce the CPU load on end node systems. Modern network adapters and operating systems have a number of features to improve efficiency, even on standard-size packets. I'd suggest testing your actual throughput without jumbo frames and see what you get. Using iperf, I can get single-thread performance at wire speed (9.89Gbit/sec on a 10GbE link). If you need to use multiple iperf threads to get wire speed, you may benefit from enabling jumbo frames.

Most protocols should correctly handle the case of a session between a system configured for jumbo frames and one that isn't. Where people tend to run into problems is where both end nodes are configured for jumbo frames, but something on the path between those two nodes doesn't support jumbo frames. This results in packets being "dropped on the floor".
 
  • Like
Reactions: TheBloke

Chris Buchanan

New Member
Jun 6, 2016
5
0
1
45
I see, that was what I was going to do between my gigabit switch with 10gb ports and the LB6M with LACP. Would really help with the connections between the SANs / HyperV and the desktops.

I can get about 300 - 400 MBps with no jumbo frames, so I think I will enable it to see what I get then.
 

josh

Active Member
Oct 21, 2013
611
181
43
Hello! I'm having a bit of a problem fiddling around with the management ports.

Is there any way to turn one of the management ports into a switchable Ethernet port? Also, do you SSH into the switch on the management port? I've tried configuring SSH by using the network parms command on page 7 but I can't seem to ping (much less SSH into) the configured IP from a machine connected to the same subnet set. From the arp cache of the subnet I see that the MAC address (X:X:X:X:1B:48) pointing to the configured IP is different from that showing under #show network (X:X:X:X:1B:47) OR #show mac-addr-table. Where exactly should I be plugging the cable to?

Also, to be noted is that the default VLAN has an IP on a different subnet (192.168.2.1). Is this a problem?

Thanks in advance!

#show network

Interface Status............................... Up
IP Address..................................... 192.168.1.254
Subnet Mask.................................... 255.255.255.0
Default Gateway................................ 192.168.1.1
Burned In MAC Address.......................... X:X:X:X:1B:47
Locally Administered MAC address............... 00:00:00:00:00:00
MAC Address Type............................... Burned In
Configured IPv4 Protocol....................... None
Management VLAN ID............................. 1
 

Sleyk

Your Friendly Knowledgable Helper and Techlover!
Mar 25, 2016
1,361
707
113
Stamford, CT
Well, lets see, I dont think the management ports can switch internet packets. Im not too sure though. The other ports 25-28 are switched as you already know.

You can ssh into the management ports or the regular ports as long as you set a static ip to the port. You could use the management port to setup network parms. I just gave the management port the only ip address through network parms first, then you can plug up to the other ports (25-28) and the switch lets me ssh in through the regular ports. No issues so far.

At first setup, you will have to use the single port by itself called "console" to setup the ip address on the switch.

Don't forget to use a cisco serial cable, or a usb to serial console cable. If you plug up a plain Ethernet cable to the switch on first setup, it wont work, as you have to enable the ssh server. The switch only responds to telnet at first, then, you can enable ssh on it. That's why you need the serial/usb console cable to get a "com" port to connect through telnet.

Ok so let's see. If all that is in place, did you type the command "network protocol none" after "show network"? This turns off the dhcp server, so that you can only get to the switch through static ip you will set. Then, try the network parms syntax my friend.
 
Last edited:

josh

Active Member
Oct 21, 2013
611
181
43
I followed every step in that post. I've used the console port to set up more than the management IP. I've set up VLANs and interVLAN routing. The IP on #show network is static. Does it matter that 192.168.1.1 (gateway IP) is the address of another router which serves as a gateway to the internet and also runs a DHCP server?

Also, if I connect machines to a VLAN with an IP address set, does it mean that this IP serves as the IP to the switch and can be used for SSH as well?