My experience with Cloudblast

averagedatahoarder · September 22

Have been a lurker on LET for a bit and more recently a member. I have several servers on @cloudblast so I thought I’d share my experience.

Will start with cons:

NL servers are slow, they use 7551P and noticed ~40% hypervisor steal. US/UK regions worked fine so I suspect just an issue with NL.
Support is a bit slow but faster on Discord.
First top up didn’t get credited to my account but opened a ticket and they fixed it + provided credit.
Routing was a bit cursed for me pulling data, but probably mostly my fault.

Pros:

Unlimited bandwidth is (seems?) actually unlimited, I’ve pushed nearly 500TB of traffic through in a month with no throttling.
Able to sustain 5-7Gbps up/down.
Good dashboard, nice API. Accepts crypto.
Great bandwidth to price ratio, probably the best I have found.

Overall very happy with their service (as someone optimizing for bandwidth) and running 4 servers with them. Performance on non-NL regions is good. Especially for bandwidth IMO it is a great option

averagedatahoarder · September 22

Also- had a lot of trouble getting a SLC VPS, so I setup a script that queried their API every few seconds to automatically provision something whenever it became available

itzsenu · September 22

@averagedatahoarder said: Unlimited bandwidth is (seems?) actually unlimited, I’ve pushed nearly 500TB of traffic through in a month with no throttling.

what are you running to actually use that much bandwidth?

@cloudblast if it’s truly unlimited, is it okay to use around 1 to 3 PB of bandwidth per month?

averagedatahoarder · September 22

what are you running to actually use that much bandwidth?

Mirroring Linux ISOs (jkjk)

averagedatahoarder · September 22

I guess I'd ask you @itzsenu, what are YOU hosting that needs 1-3PB of bandwidth? :P

itzsenu · September 22

@averagedatahoarder said:
I guess I'd ask you @itzsenu, what are YOU hosting that needs 1-3PB of bandwidth? :P

it’s a video streaming system where it only streams the changed pixels, with a max 1 to 2 ms delay on the server side, excluding the user’s network and location latency.

                    ┌─────────────────────────────────────────────────────┐
                    │              GLOBAL CONTROL PLANE                    │
                    │  - Session registry (distributed, replicated)        │
                    │  - Leader election (Raft/Paxos across mother nodes)  │
                    │  - Global health telemetry aggregator                │
                    └───────────────┬───────────────────────────────────────┘
                                    │
          ┌─────────────────────────┼─────────────────────────┐
          │                         │                          │
  ┌───────▼────────┐      ┌────────▼────────┐       ┌─────────▼───────┐
  │ MOTHER-REGION-A │◄────►│ MOTHER-REGION-B │◄─────►│ MOTHER-REGION-C  │  ← MOTHER TIER
  │ (Primary, Asia) │      │ (Primary, EU)   │       │ (Primary, LATAM) │    (multi-region,
  │  + hot standby  │      │  + hot standby  │       │  + hot standby   │     active-active)
  └───────┬─────────┘      └────────┬────────┘       └─────────┬────────┘
          │                         │                          │
    ┌─────┴─────┐             ┌─────┴─────┐              ┌─────┴─────┐
    │  TIER 1   │             │  TIER 1   │              │  TIER 1   │
    │  MESH A   │             │  MESH B   │              │  MESH C   │
    │ Child 1-N │◄──cross─────│ Child 1-N │──cross──────►│ Child 1-N │   (inter-region
    │ (VPS)     │  region     │ (VPS)     │  region      │ (VPS)     │    mesh sync for
    └─────┬─────┘  sync       └─────┬─────┘  sync        └─────┬─────┘    failover only)
          │                         │                          │
    ┌─────▼─────┐             ┌─────▼─────┐              ┌─────▼─────┐
    │  TIER 2   │             │  TIER 2   │              │  TIER 2   │
    │  RELAY A  │◄───────────►│  RELAY B  │◄────────────►│  RELAY C  │   (dedicated,
    │ (Dedicated│   sync      │ (Dedicated│   sync       │ (Dedicated│    active-active
    │  + backup)│             │  + backup)│              │  + backup)│    with failover)
    └─────┬─────┘             └─────┬─────┘              └─────┬─────┘
          │                         │                          │
    ┌─────▼──────┐            ┌─────▼──────┐             ┌─────▼──────┐
    │  TIER 3    │            │  TIER 3    │             │  TIER 3    │
    │  EDGE A1-N │            │  EDGE B1-N │             │  EDGE C1-N │   (geo-nearest
    │  (VPS)     │            │  (VPS)     │             │  (VPS)     │    to end users)
    └─────┬──────┘            └─────┬──────┘             └─────┬──────┘
          │                         │                          │
       Users                     Users                      Users
    (Asia region)              (EU region)               (LATAM region)

(note: this diagram was drawn by GPT, but the architecture behind it is the same.)

oloke · September 22

@averagedatahoarder said: NL servers are slow, they use 7551P and noticed ~40% hypervisor steal

In NL steal is real, I get between 15 and 20% on my node (under load).
The CPU performance is also pretty poor getting just over 300 single core GB6 score (i posted yabs today)

In the UK however, the node is actually really good as of now, no steal and performance exceeding what I would expect. (yabs)
Bit weird since in theory it's the same product, just different location. UK appears to have 4x better score. Maybe that's because there are not many customers there yet - UK was recently launched.

Support is a bit slow but faster on Discord.

Support has been helpful and quick in my case, always resolving stuff within few hours. And over past 3 days I managed to open 4 tickets

@averagedatahoarder said: Good dashboard, nice API. Accepts crypto.
Great bandwidth to price ratio, probably the best I have found.

Definitely true They also give you 2% bonus on top-ups using cryptocurrency.

averagedatahoarder · September 22

@oloke said:

@averagedatahoarder said: NL servers are slow, they use 7551P and noticed ~40% hypervisor steal

In NL steal is real, I get between 15 and 20% on my node (under load).
The CPU performance is also pretty poor getting just over 300 single core GB6 score (i posted yabs today)

In the UK however, the node is actually really good as of now, no steal and performance exceeding what I would expect. (yabs)
Bit weird since in theory it's the same product, just different location. UK appears to have 4x better score. Maybe that's because there are not many customers there yet - UK was recently launched.

Support is a bit slow but faster on Discord.

Support has been helpful and quick in my case, always resolving stuff within few hours. And over past 3 days I managed to open 4 tickets

@averagedatahoarder said: Good dashboard, nice API. Accepts crypto.
Great bandwidth to price ratio, probably the best I have found.

Definitely true They also give you 2% bonus on top-ups using cryptocurrency.

UK runs 7642 vs 7551P on NL so maybe expected?

network · September 22

@averagedatahoarder said: I’ve pushed nearly 500TB

What's your total monthly cost with them?

averagedatahoarder · September 22

@network said:

@averagedatahoarder said: I’ve pushed nearly 500TB

What's your total monthly cost with them?

Under $50/mo (slightly over counting the current NL node I have, which I'm in the process of migrating to UK node). But most traffic is coming from 2 servers each 9.6 euros a month

cloudblast · September 23

Thank you guys, glad to see positive reactions to all the work we've put behind cloudblast!

Yes, NL was our first location and our nodes there are running on EPYC 7551p's.

But all our other locations runs on either EPYC Rome 7642 (UK) or EPYC Milan 7C13, 7663 or 7713 (USA and HK)

mhpteam · September 23

@itzsenu said:

@averagedatahoarder said:
I guess I'd ask you @itzsenu, what are YOU hosting that needs 1-3PB of bandwidth? :P

it’s a video streaming system where it only streams the changed pixels, with a max 1 to 2 ms delay on the server side, excluding the user’s network and location latency.

                    ┌─────────────────────────────────────────────────────┐
                    │              GLOBAL CONTROL PLANE                    │
                    │  - Session registry (distributed, replicated)        │
                    │  - Leader election (Raft/Paxos across mother nodes)  │
                    │  - Global health telemetry aggregator                │
                    └───────────────┬───────────────────────────────────────┘
                                    │
          ┌─────────────────────────┼─────────────────────────┐
          │                         │                          │
  ┌───────▼────────┐      ┌────────▼────────┐       ┌─────────▼───────┐
  │ MOTHER-REGION-A │◄────►│ MOTHER-REGION-B │◄─────►│ MOTHER-REGION-C  │  ← MOTHER TIER
  │ (Primary, Asia) │      │ (Primary, EU)   │       │ (Primary, LATAM) │    (multi-region,
  │  + hot standby  │      │  + hot standby  │       │  + hot standby   │     active-active)
  └───────┬─────────┘      └────────┬────────┘       └─────────┬────────┘
          │                         │                          │
    ┌─────┴─────┐             ┌─────┴─────┐              ┌─────┴─────┐
    │  TIER 1   │             │  TIER 1   │              │  TIER 1   │
    │  MESH A   │             │  MESH B   │              │  MESH C   │
    │ Child 1-N │◄──cross─────│ Child 1-N │──cross──────►│ Child 1-N │   (inter-region
    │ (VPS)     │  region     │ (VPS)     │  region      │ (VPS)     │    mesh sync for
    └─────┬─────┘  sync       └─────┬─────┘  sync        └─────┬─────┘    failover only)
          │                         │                          │
    ┌─────▼─────┐             ┌─────▼─────┐              ┌─────▼─────┐
    │  TIER 2   │             │  TIER 2   │              │  TIER 2   │
    │  RELAY A  │◄───────────►│  RELAY B  │◄────────────►│  RELAY C  │   (dedicated,
    │ (Dedicated│   sync      │ (Dedicated│   sync       │ (Dedicated│    active-active
    │  + backup)│             │  + backup)│              │  + backup)│    with failover)
    └─────┬─────┘             └─────┬─────┘              └─────┬─────┘
          │                         │                          │
    ┌─────▼──────┐            ┌─────▼──────┐             ┌─────▼──────┐
    │  TIER 3    │            │  TIER 3    │             │  TIER 3    │
    │  EDGE A1-N │            │  EDGE B1-N │             │  EDGE C1-N │   (geo-nearest
    │  (VPS)     │            │  (VPS)     │             │  (VPS)     │    to end users)
    └─────┬──────┘            └─────┬──────┘             └─────┬──────┘
          │                         │                          │
       Users                     Users                      Users
    (Asia region)              (EU region)               (LATAM region)

(note: this diagram was drawn by GPT, but the architecture behind it is the same.)

wow, our gpts have same mind, my infrastructure is 90% the same as this.

vbloods · September 23

sounds dodgy ...

averagedatahoarder · September 23

No longer have 4 VPS's, nothing wrong abt Cloudblast but can't afford 4 💀 and was able to optimize my apps to consume significantly less memory. Let's see if I'm still able to push 500TB through next month

averagedatahoarder · September 23

Also an update, am now at 3TB on UK region

Howdy, Stranger!

Categories

In this Discussion

My experience with Cloudblast

Comments

Howdy, Stranger!

Quick Links

Categories

In this Discussion

My experience with Cloudblast

Comments