Announcement

Collapse
No announcement yet.
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Network slowness?

    So.... Since we've had our Shoretel system installed almost 2 yrs ago now, I have had so many complaints about our network being slow (and verified the slowness), I am wondering how many here have the same type of setup I do?

    OLD SETUP:
    100MB flat network with all building (4 total) connected to LAN via GB fiber.

    NEW SETUP (Shoretel):
    100MB VLAN segmented network. Each building with their own VLAN connected still with GB fiber. All PC's connect through Shoretel 230 IP phone.


    Is it common practice to combine both data & voice networks and is this the way you would of done it?

    I am really starting to think that maybe I should seperate the networks.

    Thoughts?

  • #2
    Slowness

    According to your descriptions, you shouldn't have noticed any real difference in the network performance from before to after.

    when you say slowness, do your ping times actually increase during the slowness?

    The PCM application is a real resource hog. Is it possible all your machines are running slower (out of memory) and it just *seems* like a network issue? If you are running outlook in cached mode and using the PCM, it can bring Client PC's to their knees.

    Comment


    • #3
      Pings seem to be about the same that I can tell and my PC is not running Outlook in cached mode. I also have a quad core wth 4GB RAM so I don't think that is the issue.

      Comment


      • #4
        Ping Times

        what are your ping times roughly?

        A quad core with 4GB of ram can absolutely be crushed by the PCM if running in cached mode. Here is how:

        1. The PCM keeps the outlook OST file open
        2. Outlook tries to close and cant really close the OST
        3. The next time outlook opens, it initiates a background scan on the ost file. This scan causes the entire OST to be checked and scan. On a large (1+GB ost) file, this can take more than an hour. During this entire time the pc's hard disk is monopolized by outlook, causing the entire pc to run like crap.
        4. The users get frustrated at the performance, and restart outlook.
        5. Outlook restarts the OST scan
        6. the cycle continues

        make sure an exclude this as a possilbe cause before you dig too deep into anything else.

        Comment


        • #5
          What do you want me to ping? When I ping one of my servers it stays at or around 2ms. The only one that I think is out of line is when I ping my Sonicwall firewall. That jumps anywhere from 20ms to 80ms. Seems high to me.

          Comment


          • #6
            ping

            Pick a server that gets a lot of performance complaints since the shoretel system was installed. start a continuous ping. open some files/data off of that server and see if the times start spiking.

            i would also leave the ping running for a long time, like a whole day. at the end of the day you can stop it and see what your highest ping time was. you may have periods of the day where the 100mb cant keep up.

            you could also have something as simple as a duplex mis-match between the phones and the pc's (but be ok from the phone to the switch). it would cause PC traffic to be terrible, and yet phone traffic be fine.......

            Comment


            • #7
              So as for my setup, are you comfortable with the way I have described it or would you look to change it? I really would like to goto a GB network but I cannot justify swapping out all the switches and phones for that since this equipment is only 2 yrs old.

              Comment


              • #8
                Setup

                Personally, I am a fan of seperate networks. Gigabit straight to the desktops, with a separate 100mb cable for the phones on their own switches.

                I would not want to run my pc's through 100mbit at all, much less THROUGH a cheap-o phone switch.

                That being said, you shouldn't have noticed a difference from BEFORE and AFTER since your network really didnt change that much.

                If the change was drastic, then something must be misconfigured somehwere.....

                Comment


                • #9
                  It changed drasictally I think from a flat topology (everything on one segment) to a 4 segment VLAN.

                  HQ (City Hall) = VLAN1 (35 PC's/Phones)
                  Community Center = VLAN2 (7 PC's/Phones)
                  Public Works = VLAN3 (15 PC's/Phones)
                  Water Treatment Plant = VLAN4 (5 PC's/Phones)

                  Double the devices now when you add in the phones and I think this is drastic personally.

                  My problem is for me to seperate them I would have to upgrade my old CAT3 to CAT5 for the phones since currently each office only has one CAT3 & one CAT5 connection.

                  Comment


                  • #10
                    Cat5

                    What kind of switches, etc are you using? Make sure the phone and PC are both running full duplex.

                    I would get a copy of etherpeak or wireshark going to look for problems before you change anything. You could just have a bad switch causing problems.

                    Do your speed problems seem to come and go, or does it seem slow all the time?

                    I have seen a situation once where someone applied QOS a bit incorrectly. They had it set in such a way where all traffic was limited all the time, not just when phones were in use.

                    can you post your VLAN and QOS configs?

                    I have also had very good luck splitting single cat5 runs into two separate drops, removing the need for additional runs. I dont think I would do this on a mass scale, but it does seem to work very well. Panduit data jack ends seem to be the best for this.

                    Comment


                    • #11
                      Switches are Adtran NetVanta 1335 POE, 1524ST & 1224ST POE's. Speed problems come and go.
                      How do I post VLAN/QOS configs?

                      Comment


                      • #12
                        Qos

                        Log into your netvanta switches using a web browser and their IP address. If they dont have one, you need a special proprietary serial adtran console cable.

                        Go to Class of Service and attach that screenshot.

                        also, check and make sure your ports all list 100/full for their status. make sure your gig uplinks list 1000 full on their status

                        if your problems come and go, I would think you are just bumping into 100Mbit ethernet restrictions.

                        Are all your switches in each individual building stacked with 1000mbit as well, not just the fiber lines between the buildings?

                        Take your biggest complainer, split her/his cable into two drops at their desk and in the data room. Plug their PC data cable directly into one of the adtrans 100mbit ports, see if their problem goes away. This removes the phones as a cause. If the problem goes away, then you have isolated the issue to the phones and thier internal switches, etc. If the problem doesn't go away, plug their PC data cable into one of the 1000mbit ports on the adtran. If this fixes the problem then you know that 100mbit just isnt enough for you any more.


                        If none of the above fixes the problem, then you have a routing/vlan/qos/physical cabling problems. This is because you ruled out the phones, the 100mbit limitation problem, the phone switches, etc. I would run for several days in each configuration setup to rule out "coincidences"

                        Comment


                        • #13
                          I have used my PC as a test for some of this (including this morning) and my PC is not connected through my phone and I am running GB network on mine and the problem reported this morning also happened on my PC. All servers are connected to the 1524ST which is a GB switch and all switches regardless of being 100MB or 1GB have all ports set to full duplex.

                          See attached for COS.
                          Attached Files

                          Comment


                          • #14
                            Problem

                            What were the symptoms of the problem exactly? How long did the slowdown last? When the network slowed down, I assume it was for many different things, it wasnt isolated to data to/from one particular server correct? The fiber between your buildings plugs directly into the adtrans right, no firewall/router? If your ping times are not rising during the network "slowdown" period, then it is usually NOT network related at all. The ping times tell you that your data is getting from point to point quickly without delay.

                            I requested a screenshot from someone i know that has Adtran QOS setup to compare. Although, if your QOS was wrong, it would affect the phone quality more than anything, and you arent having any phone quality issues..........

                            From all the information , it almost sounds like a broadcast storm type situation/switching loop. I had a client recently that had something similar. They had a small switch in a conference room that they plugged two data drops into. This caused a switching loop (since Spanning Tree has to be disabled for the shoretel to be happy) and caused the entire network to veg out. When they used the room, the would plug their laptop in and any cables laying on the table. When they left they would unplug. One of the cables came out of the switch, and they plugged it back into the switch..........

                            Run wireshark on your pc and do a snif when the network is freaking out.... see if there are a crazy amount of broadcasts going on.

                            Without spanning tree on, any switch that is connected back into itself, or any loops or duplicate uplinks will cause the entire network to act very badly.... it is normally more consistent though.

                            Comment


                            • #15
                              I am running almost the exact same setup, 55 offices using 1335 POE's and 1224POE switches. One thing I would do is set dscp46 to match COS 6. I had similar problems with larger office until I did this.

                              On the screen shot you showed there is a place at the bottom of it to put it in. I use the CLI so Ive never seen that screen, however this is what it looks like when consoled into the 1335:

                              qos cos-map 1 0 1
                              qos cos-map 2 2 3
                              qos cos-map 3 4
                              qos cos-map 4 5 6 7
                              qos queue-type wrr 1 2 3 expedite
                              !
                              qos dscp-cos 46 to 6

                              This gives COS 5,6 and 7 expedite status, and when it sees dscp 46 (Shoretel Calls) it marks them as COS 6 .

                              Comment

                              Working...
                              X