Howdy, Stranger!

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


Do good providers balance at the node level?
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.

Do good providers balance at the node level?

For such a scenario:

If on the same node, everyone's vps has high traffic during London Time 6-10 pm, will AWS, Azure, Hetzner, Netcup, etc automatically transfer some vpses to other nodes to reduce the rush hour drama, like we do teleworking, flexible working schedule, car pool, to reduce or avoid the traffic during rush hours?

Just curious, because if there is no balancing, the node will freeze or congest during those rush hours, unless high level of overhead redundancy is available, which almost guarantee higher price tag.

Comments

  • @letlover said: automatically transfer some vpses to other nodes to reduce the rush hour drama

    why would they do that?

    put a 10gig in hypervisor/node and call it a day. or maybe 100gig NIC.

    if a particular vm is using majority of the bw, then throttle them down

    Thanked by 1letlover
  • letloverletlover Member
    edited July 2022

    @nanankcornering said:

    @letlover said: automatically transfer some vpses to other nodes to reduce the rush hour drama

    why would they do that?

    put a 10gig in hypervisor/node and call it a day. or maybe 100gig NIC.

    if a particular vm is using majority of the bw, then throttle them down

    Probably forced throttling is the easiest way to reduce the rush hour traffic.

    Then if that vps only uses 2-4 hours everyday but only during rush hours, seems throttling will cause a lot of customer reactions. :)

  • @letlover said: the node will freeze or congest

    What do you mean by that? Even if all bandwidth is utilized and netlink saturated, it won't affect the
    node itself or the CPU. The big ones normally have link aggregation to the switch so they have quite a lot of redundancy for such cases. The smaller providers are the ones who usually get affected by this, but also depends if they oversell. Obviously if it's a $10/year deal, with 100 VMs
    on a node and everyone is "promised" a shared 1gbit link, things will get rough.

  • yoursunnyyoursunny Member, IPv6 Advocate
    edited July 2022

    To reduce rush hour congestion:

    • road: increase dynamic toll prices
    • airport: make you wait in the queue
    • cloud compute: terminate spot instances
    • data center network: drop traffic of lower tier customers
    • Internet exchange: drop traffic of lower tier customers
    • cellular network: throttle discounted lines and traffic identified as streaming video
    Thanked by 2letlover lanefu
  • @luckypenguin said:

    @letlover said: the node will freeze or congest

    What do you mean by that? Even if all bandwidth is utilized and netlink saturated, it won't affect the
    node itself or the CPU. The big ones normally have link aggregation to the switch so they have quite a lot of redundancy for such cases. The smaller providers are the ones who usually get affected by this, but also depends if they oversell. Obviously if it's a $10/year deal, with 100 VMs
    on a node and everyone is "promised" a shared 1gbit link, things will get rough.

    If every vps reaches its peak during those rush hours, cpu and ram and disk i/o may all reach peak performance. Yet, vpses are generally oversold, about 2 to 3 times more than the cpu and ram, so it may freeze the node if every vps is at the turbo mode.

  • @yoursunny said:
    To reduce rush hour congestion:

    • road: increase dynamic toll prices
    • airport: make you wait in the queue (buy TSA Precheck to jump the queue)
    • cloud compute: terminate spot instances
    • data center network: drop traffic of lower tier customers
    • Internet exchange: drop traffic of lower tier customers
    • cellular network: throttle discounted lines and traffic identified as streaming video

    how can we as end user elevate our tier? probably use money. hope there are some cheaper ways. :)

  • @yoursunny said:
    To reduce rush hour congestion:

    • road: increase dynamic toll prices
    • airport: make you wait in the queue (buy TSA Precheck to jump the queue)
    • cloud compute: terminate spot instances
    • data center network: drop traffic of lower tier customers
    • Internet exchange: drop traffic of lower tier customers
    • cellular network: throttle discounted lines and traffic identified as streaming video

    It is like fast lanes vs slow lanes, and there are car pool lanes, and emergency lanes, on highway. We mundane joes at LET probably mostly stay in slow lanes.

  • yoursunnyyoursunny Member, IPv6 Advocate
    edited July 2022

    @letlover said:
    how can we as end user elevate our tier? probably use money. hope there are some cheaper ways. :)

    • road: pay toll; break your leg and get handicapped plate
    • airport: buy TSA Precheck
    • cloud compute: create standard instances instead of spot instances
    • data center network: buy OVH instead of SoYouStart
    • Internet exchange: peer with more networks
    • cellular network: buy Verizon postpaid instead of Red Pocket prepaid
    Thanked by 3letlover lanefu Erisa
  • @yoursunny said:

    @letlover said:
    how can we as end user elevate our tier? probably use money. hope there are some cheaper ways. :)

    • road: pay toll; break your leg and get handicapped plate
    • airport: buy TSA Precheck
    • cloud compute: create standard instances instead of spot instances
    • data center network: buy OVH instead of SoYouStart
    • Internet exchange: peer with more networks
    • cellular network: buy Verizon postpaid instead of Red Pocket prepaid

    buy,buy,buy, oh,yeh.

    Thanked by 1yoursunny
  • so there is a reason why aws, azure, gc, etc are more expensive, as they have more redundancy and more overhead.
    not sure where hetzner, netcup, php-friends, liteserver, hybula stand when comparing to these first tier players.

  • lanefulanefu Member

    @letlover said:
    so there is a reason why aws, azure, gc, etc are more expensive, as they have more redundancy and more overhead.
    not sure where hetzner, netcup, php-friends, liteserver, hybula stand when comparing to these first tier players.

    It's the difference between using your drink coupons and staying within the resort premises vs staying at the motel across the across street and finding your own trouble.

  • @lanefu said:

    @letlover said:
    so there is a reason why aws, azure, gc, etc are more expensive, as they have more redundancy and more overhead.
    not sure where hetzner, netcup, php-friends, liteserver, hybula stand when comparing to these first tier players.

    It's the difference between using your drink coupons and staying within the resort premises vs staying at the motel across the across street and finding your own trouble.

    Yeh, rich vs poor. As always.

  • jarjar Patron Provider, Top Host, Veteran

    This is definitely one of those things you'd default to no on, and only accept a different answer if explicitly stated by the provider.

    Thanked by 1letlover
  • jarjar Patron Provider, Top Host, Veteran

    @letlover said:
    so there is a reason why aws, azure, gc, etc are more expensive, as they have more redundancy and more overhead.
    not sure where hetzner, netcup, php-friends, liteserver, hybula stand when comparing to these first tier players.

    Get 10 VPS across a wealth of reasonably recommended providers, including the top clouds like AWS and Azure. Compare uptime over 10 years. The difference will never justify the big name clouds.

    The big name clouds are at their best by providing you with all of the solutions you need to scale on demand. A strong API, object storage, block storage, database as a service, all that jazz. They're not objectively better, they're just more prepared for customers who need to be able to scale their apps on demand.

    Thanked by 2letlover yoursunny
  • @jar said:

    @letlover said:
    so there is a reason why aws, azure, gc, etc are more expensive, as they have more redundancy and more overhead.
    not sure where hetzner, netcup, php-friends, liteserver, hybula stand when comparing to these first tier players.

    Get 10 VPS across a wealth of reasonably recommended providers, including the top clouds like AWS and Azure. Compare uptime over 10 years. The difference will never justify the big name clouds.

    The big name clouds are at their best by providing you with all of the solutions you need to scale on demand. A strong API, object storage, block storage, database as a service, all that jazz. They're not objectively better, they're just more prepared for customers who need to be able to scale their apps on demand.

    So as we LET customers with no need for sophisticated requirements, actually we are at disadvantage to use these big players. To a certain degree, I agree with you.

    Thanked by 1jar
  • lanefulanefu Member

    @jar said:

    The big name clouds are at their best by providing you with all of the solutions you need to scale on demand. A strong API, object storage, block storage, database as a service, all that jazz. They're not objectively better, they're just more prepared for customers who need to be able to scale their apps on demand.

    You're also paying for a metric shit ton of checkbox security and turn-key compliance.

    If you have any customer contracts that have particulars..... The bulk of the work to prove your house is order is done.

    https://aws.amazon.com/artifact/

    Thanked by 2jar letlover
  • tjntjn Member

    On the flipside, I've always wondered how low-end hosts do it.
    I know each provider is different, but do the majority use RAID for example in case an HDD/SSD fails?

    Thanked by 1lanefu
  • jarjar Patron Provider, Top Host, Veteran

    @tjn said:
    On the flipside, I've always wondered how low-end hosts do it.
    I know each provider is different, but do the majority use RAID for example in case an HDD/SSD fails?

    Majority use RAID is probably a safe statement, but if it's not stated, always ask.

    Thanked by 1tjn
  • letloverletlover Member
    edited July 2022

    @tjn said:
    On the flipside, I've always wondered how low-end hosts do it.
    I know each provider is different, but do the majority use RAID for example in case an HDD/SSD fails?

    For storage vps, I paid special attention. Liteserver is raid 10, all others, hosthatch, inception hosting, etc, use raid 5 or raid 6. For general vps, I see clearly that php friends, netcup, hetzner have raid 10.

    Thanked by 2yoursunny tjn
  • NeoonNeoon Community Contributor, Veteran

    You can buy/reserve bandwidth allocation to avoid this, but it does not look like an issue to me if the provider takes care that no one abuse the gigabit port 24/7.

    Also, recently we see a lot of nodes with 10gig uplinks, so should become less of an issue.

    Thanked by 1letlover
Sign In or Register to comment.