Howdy, Stranger!

It looks like you're new here. If you want to get involved, click one of these buttons!


help me understand port speed
New on LowEndTalk? Please Register and read our Community Rules.

All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.

help me understand port speed

vanarpvanarp Member
edited March 2013 in Help

1Gbps port
100Mbps port

What this specification exactly indicates in the VPS Offers?

«1

Comments

  • @vanarp said: 1Gbps port

    = network speed a theoretical maximum of 125 megabytes/second

    @vanarp said: 100Mbps port

    = network speed a theoretical maximum of 12.5 megabytes/second

    Trolling a VPS provider for not giving guaranteed network speed will not make friends. Instead, most (all?) of us guarantee that you will be able to use 100% of your monthly transfer allocation.

  • shovenoseshovenose Member, Host Rep

    What @Damian said is spot on.
    That simply means the speed at which the server is connected to the internet.

  • wdqwdq Member

    It's usually the port speed/internet speed of the VPS node/server. A single server may have dozens of different people on it all using a little bit of that port speed. A 1Gbps port is almost always better considering 100Mbps isn't very much nowadays.

    Some hosts are different and say something like 50Mbps port and by saying that they mean 50Mbps per VPS meaning you'll be able to use that 50Mbps. If it's 1Gbps for the whole node you won't be able to use all 1Gbps at any given time in most cases.

  • shovenoseshovenose Member, Host Rep

    @wdq 100Mbps is plenty for most VPS nodes unless it's either incredibly oversold or there are one or more abusers on the node.
    We limit the users of our smallest plan to 50Mbps to prevent a cheap VPS customer causing issues for the whole node.
    But the node itself is still on a 100 meg port.

  • DStroutDStrout Member
    edited March 2013

    Just to be clear: 100mbps means 100 megabits per second. Since a bit is one eighth of a byte, you get one eighth of 100 megabytes per second, which equals 12.5 megabytes per second. Likewise with 1gbps.

  • @Damian said: Trolling a VPS provider for not giving guaranteed network speed will not make friends.

    The question is for my learning sake only.

    Now that advertised speed is for the VPS or the Physical Node containing the VPS?

    I know there are 100mbps & 1gbps Ethernet ports. But that should be at the node level. What if there are some 20 VPS on that node and they all share the connection, was the question in my mind.

  • @vanarp said: Now that advertised speed is for the VPS or the Physical Node containing the VPS?

    They usually clarify. Most of the time its talking about the container I believe.

  • @HalfEatenPie said: They usually clarify. Most of the time its talking about the container I believe.

    It has not been very clear from most of the offers I went through here on LET.

  • @vanarp said: It has not been very clear from most of the offers I went through here on LET.

    Then contact the provider and ask for clarification. This is communication and they're not doing their job properly by not clarifying if its per VPS container or the entire node (with each container being limited or unlimited).

  • @shovenose said: 100Mbps is plenty for most VPS nodes unless it's either incredibly oversold or there are one or more abusers on the node.

    Yes, when you have a suddenly popular large file, like a new software release or maybe a video that's been linked from a popular site, or any other peak demand event, just pop a notice on your website indicating people should download one at a time, and to form an orderly queue.

    Don't worry, vendors know best!

  • Another thought I was having before asking this question was that the providers might be physically keeping multiple (close to number of VPS if not equal ) Ethernet ports to ensure the speed advertised.

  • @vanarp said: What if there are some 20 VPS on that node and they all share the connection, was the question in my mind.

    Shouldn't be that noticeable, especially with the bandwidth allocations given for these offers versus the amount of transfer a gigabit connection can handle versus containers that aren't in use versus container usage, etc.

    For example:

    This node has 37 clients on it:
    image

    This node has several hundred clients on it:
    image

    Proper monitoring and management by the host is paramount.

  • @Damian said: Proper monitoring and management by the host is paramount.

    I totally agree. What could have caused the spike in first chart? Are the hosts throttled to certain Mbps in the second chart?

  • MaouniqueMaounique Host Rep, Veteran
    edited March 2013

    For example:
    1. We offer for all regular VPS 1 Gbps port speed (node connection) on a fair-share basis.
    2. Storage plans have 100 mbps speed BUT, KVM ones have 100 mbps limit each on a node with 1 gbps port.
    This is because backups can go VERY fast in the same DC and will probably saturate the port on the node hosting the VPS which saves the back-up.
    100 mbps should be enough since it can use the full speed because the 1 Gbps port allows 10 full speed backups at 100 mbps. If all the port speed was 100 mbps it is possible that more backups run same time and will have to squeeze through the same small pipe. With port speed 1 Gbps and limit each VM to 100 mbps you make sure there is very little chance that the port is saturated while providing decent speeds for backups. Also, since the space can be mounted and the backup saved directly, there is also not that big io activity on both storages.
    We advertise 1 gbps port speed for the KVM backup and say each vm can use 100 mbps of that (they are limited at the router level) but the physical port is still 1 Gbps.
    There is also the case where the port speed is 1 Gbps, but the VPS company only has a line of 100 mbps from the datacenter. Then, it is obvious only 100 mbps will go to and come from the Internet, but locally there might be higher speeds, for example, for back-ups, then there will be no limitation of the speed between servers in the same rack and can go as fast as the hardware (storage and NIC) allows.
    Also, these speeds are usually duplex, which means there can be 1 gbps flows in both directions simultaneously, effectively being a 2 gbps connection, since in the same time you can send and receive 2 gbps of data.

    TL;DR Port speed <> connection speed (BandWidth) for the VM. The internet connection speed can be limited in various ways and also varies with distance and routing while the port speed is usually fixed in HW.

  • @vanarp said: What could have caused the spike in first chart?

    The green chart/positive numbers are "in" to the server, so someone might have been uploading something to their VPS.

    The hosts in the second chart are not throttled.

  • @Maounique said: they are limited at the router level

    Are you sure the throttling is done at router but not the switch?

    @Damian said: The hosts in the second chart are not throttled.

    Is that a shared hosting node or OpenVZ node?

  • @vanarp said: Is that a shared hosting node or OpenVZ node?

    Probably OpenVZ (from my speculation)

  • MaouniqueMaounique Host Rep, Veteran

    @vanarp said: Are you sure the throttling is done at router but not the switch?

    It can be done in both places, if it is at switch, it is usually limited by the port speed (physical one). If the switch and the node ports are different, then the lower speed one sets the exchange rate, but it is silly to do that, nobody will introduce physical bottlenecks, will rather rate-limit and do QoS in software or in the core router for fast lines (10 gbps and up).

  • Here is one of our full nodes;

    And a near empty node, with one outbound DoS incident that we dealt with.

  • By the way @SimpleNode and @Damian, are those Observium graphs?

  • @vanarp said: Is that a shared hosting node or OpenVZ node?

    OpenVZ.

    @HalfEatenPie said: are those Observium graphs?

    Yesh

  • SimpleNodeSimpleNode Member
    edited March 2013

    @HalfEatenPie said: Observium

    Yes.

    I love the temperature graphs;

  • @SimpleNode said: I love the case temperature graphs;

    was that some maintenance downtime indicated by the brief missing of line on first Saturday?

  • @SimpleNode said: I love the temperature graphs;

    Some of ours work!

    image

  • @Damian said: Some of ours work!

    wow that looks like a cityscape outline :)

  • @SimpleNode and @Damian: Me gusta

    @vanarp said: was that some maintenance downtime indicated by the brief missing of line on first Saturday?

    This can be anything from Network issues, downtime, or just Observium not being able to communicate with SNMP on that server for a while.

    @Damian: Seems to be irregular temperatures?

  • MaouniqueMaounique Host Rep, Veteran

    @Damian said: Some of ours work!

    While others flatline :P

  • @Jarland setup our Munin monitoring system so I don't know if we even have temperature sensors on ours...

  • @Damian said: Some of ours work!

    Our ones work too! 29.7C to 29.8C. Stable temperatures are good... right?

    I believe these are case temperatures not an actual component, so that's why they're so stable.

  • nice topic, thanks all for detailed information..

Sign In or Register to comment.