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.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Excited about it. Keep in mind this is network storage that mounts as a true, local volume. This isn't intended to be compared to backup space or object store. Though, it certainly could be used for storing backups.
Would have been nice if they emailed those who applied to the beta program and never got accepted to let them know.
I was hoping it would be a bit cheaper, hopefully though the IOPS are high as a result of the average level pricing. Might be useful for some extra storage with Elasticsearch. Gotta wait for it to get to our deployed region (AMS2) though...
Right, it's not the same as S3. However, when anyone quotes a per-GB price for disk, that's the first thing I (and I think most people) think about as a price benchmark.
I actually like this offer, though hope the cost comes down over time. One thing I don't like about most VPS companies is that there are only standard T-shirt sizes. I understand the economics of why, but as a consumer, I would love to be able to get a bunch of legos and put things together the way I want, and this is a step towards that.
Besides, what kind of LET member would I be if I didn't whine about cost? :-)
You can create volumes of custom size, was an excellent choice I think. Could've really overshadowed the success by leaving that out, IMO.
Oh definitely. I don't expect it to appeal to bargain hunters here or people who are just looking for the cheapest storage to toss some data in. That includes me in most situations. Certainly that wasn't the target with the product. We're talking about data mounted as a local volume, stored on SSD, replicated over multiple systems in multiple racks. Frankly, I'm surprised it came out this cheap. But this is what you build SaaS on, not a replacement for existing SaaS
That's nice!
It's LIVE!!!!! 10 cents / gb
No, everyone pays 0.10/gb
Very expensive I think.
Digital Ocean is 10 cents, Amazon is 10 cents, Google is 17 cents, Vultr is 10 cents
(Per GB per Month)
Why do you say that Digital Ocean is "very expensive"? They seem to be similar pricing to their main competitors.
Backblaze b2? 0.005 usd
Backblaze is not a Digital Ocean competitor, they are not a hosting provider, only a backup provider. You can't really compare those two things. Just like you can't compare Digital Ocean's block storage pricing to dropbox or crashplan or a SSD hard drive from Best Buy.
You really wouldn't be making the most cost effective choice to use DO's block storage as a simple copy of a large amount of personal backups (IMO). It's like adding a physical disk to your VPS. It's SSD storage replicated across multiple systems in multiple racks.
Backblaze is incredibly great. But it's comparing software to infrastructure. Two incredibly different things.
Imagine our block storage as what you build your backup company on, rather than what you spin up to backup your movies to.
What sort of IO can the block storage do 24/7?
Test it out, be sure to share feedback via ticket if the IO isn't meeting your needs for a particular application. It's not going to provide great dd benchmark porn, if that's what you're asking. It should be more than sufficient, but this is where we'll see if that remains true with people using it to a reasonable degree. Certainly if improvements need to be made later, that's not a discussion I believe anyone is opposed to having. Good to get feedback for just such a discussion.
Notice I'm avoiding hard numbers and I'll continue to do so. I don't want reactions to perfectly setup benchmarks on a clean setup, I want feedback from people using it to say "This does or does not meet my needs, here's how and why."
...though I wish they had a Linux backup client.
One of the things in my mind that separates DO, Vultr, AWS, Azure, etc. from generic hosting providers is that I can abuse IO all I want. Underneath I may be throttled from an IOPS or throughput perspective (in fact, I'm sure I am) at the VM or storage layer but if I run hard, I get consistent performance and won't have my VM turned off for abuse.
At least, I don't think I will on DO. Definitely not on AWS or Azure.
Seriously, I can't make heads or tails of getting B2 to work on Linux. It can be done, but not by me apparently. Would love to backup MXroute to it.
You'd have to pull some incredible acrobatics to do it. I won't say I haven't seen someone do it, but at this point I've pretty much seen everything at least once I think
Honestly, I expected it to be cheaper. AWS offers it also for ~0.10$ (based on region), but with that you also get a ton of features which DO is lacking aaand crazy bandwidth cost.
If your application is not bandwidth intensive I don't see any sensible reason to not use AWS.
Does anyone here have similar low traffic servers and are hosting with DO? What features of DO you are utilizing that would cost a lot more on AWS?
To be honest, I can't predict my needs down to the fine details for which AWS charges. I always wanted to love AWS, because I love Amazon and they're the clear market leader, but I don't know what my bill is going to be and that scares the mess out of me.
I have a larger reason for preference now than just the above, given my employment, but that sentiment is one you can find me holding well before I sent a resume. It's why I told Ben in my interview with him that I thought DO had a golden opportunity and needed only to grab it.
I applied for the product beta, never got the invite but apparently even people that just applied get free access until august 31st
What "ton of features" do you get with 0.10/gb at AWS? That is the storage and IO only (just like with DO). The 0.10 at AWS doesn't include snapshots, transfers, compute, etc.
The high bandwidth cost of AWS is one reason I moved to DO. 0.09/gb bandwidth out at AWS is a hugh variable each month.
At DO, you get compute + 3TB bandwidth + 40 gb SSD + 2nd IP address all for the same price as just the comparable compute at AWS. Then, you have to add the storage + bandwidth + ip address at AWS. So, DO ends up being much cheaper.
AWS does not have KVM, making it much much harder to fix a firewall mistake, etc.
As did I... won't be using this product sadly. Can't afford for personal use
I don't know what SSD drives they use, and I can only see pricing for non-enterprise Samsung SSD drives, so I will use that pricing for the calculation. 2TB Samsung PRO 850 drive = $900.
That is $0.44 per GB. Then, they have your data redundant across multiple drives on multiple racks. I don't know how many times the same data is redundant, but let's guess that your data is on 2 drives per rack, and on 2 racks. That is 4 drives holding your same data redundant.
Now, let's assume there is some compression going on, so let's calculate with 2x compression.
So, a 100GB volume is costing you $10 per month.
That is 4 drives redundant with 2x compression, so your data is using a total of 50GB per drive = 200GB
This means that DO is paying 0.44 per GB. 0.44 x 200GB = $88. DO is paying a total of $88 to hold your data, and you are paying $10 per month.
It takes DO 8 months to break even on your data, and then every 5 years or so, those drives need replaced and that 8 months starts over.
Add on top of that, this is not considering the infrustructure/network that the drives connect to, the racks, the raid cards, the servers, etc...
And, they are probably not sold out to 100% capacity, so they are not actually being paid for every GB that they have available.
Looking at it that way, 0.10 per GB does not seem as expensive
Let me know your opinions of this................
Or they use a SAN.
Well looking at it that way, if GM charged $100K for a Chevy, it would not seem expensive, but the real benchmark is the market.
To take only DO's cheapest offering, you won't get something in the same class at AWS for for $5/mo. AWS is great for some use cases but if you don't need the ton of features, it's overpriced.
"Your data is stored on hardware separated from your Droplet, and it’s replicated multiple times across different racks, reducing the chances of data loss in case of hardware failure."
https://www.digitalocean.com/features/storage/
The market for the big providers is around 0.10 for hosting storage.
This price is not for LET.
We've kind of never been, just have one little tiny entry server that qualifies.
What's the avg IO latency like? 500ms? less?
$1/10gb-month seems fair for replicated SSD.
I'll give it a try when FRA1 comes up.
Honestly, excluding bandwidth you can get a no upfront reserved t2.nano instance for about 3.4$ per month, add in block storage (gp2) for 20GB, that gives you ~5.4$ per month without including bandwidth.
The thing is, you can never be sure what you will need, for example down the road I might require scalable setup. You can easily scale vertically by just changing the instance type or you can scale horizontally by automating deployments and putting them behind ELB.
There is no scaling in DO AFAIK.
AWS offers a ton of features which allow to circumvent such issues, for example, I have yet to experience need to set up iptables on an instance, everything is handled by security groups. I have had an issues configuring sshd, which have ruined the accessibility to server, in such case I detached the volume from the inaccessible server and resolved the issues by mounting the volume on another instance and editing the required files, reattache it back to the old instance and voila.
I was not really speaking about the storage features, but okay, lets limit it to the storage. Currently as far as I know DO offers undisclosed amount of IO and no way to know if you are near any limits, there is no guaranteed IO (please correct me if I'm wrong). On AWS I am able to create cold storage drives for 0.025$ GB/Mo if I have non IO intensive workload or if I see that I'm hitting IO limits I am able to scale the IO limits by using provisioned IOPS storage.
Merged the two threads into one. Sorry, @jarland, but DO is (fortunately) not Kimsufi. ;-)