[juner]

On occasion, I have to build custom storage servers for my data center. They are usually for archive data that doesn't get accessed that often, but retention requirements make it so the data must be around for 10 years.

I am currently building the largest capacity single storage server I have built built to this point. Sourcing everything from Newegg Business, as usual. It will have two RAID 6 arrays of 8 drives with 8TB each. That translates to just over 87TB of useable space, leaving room for one more array (~43.7TB) should I need to expand in the future. Other specs are two Xeon E5's, 32GB RAM, 4U chassis with dual 1400W PSUs and 10Gb NICs. Just over $17K for the initial build. Should be fun times.

[jroa]

Make sure you put Lunix on it. 

[juner]

Make sure you put Lunix on it.

A Linux or BSD distro would be fine for this server's use case. It is just going to host some CIFS shares that Windows servers need to access. It doesn't have to run any of the proprietary Windows apps that most of my other servers do.

[Rushy]

If it is only for storage, why does it need two processors?

[juner]

If it is only for storage, why does it need two processors?

It doesn't.

[Thork]

Are you going to let people play games on it?

[juner]

Are you going to let people play games on it?

I could but it wouldn't be efficient since it lives in a data center.

[Thork]

Seems like a waste.

[juner]

Seems like a waste.

It basically is a waste. It is where old files go to die, but are on life support for about 10 years before they can be purged. Still enjoyable to put together, though.

[Fortuna]

I bet government agencies do shit like this all the time, wasting money on overblown systems just for shits and giggles. What a jackass move.

[juner]

I bet government agencies do shit like this all the time, wasting money on overblown systems just for shits and giggles. What a jackass move.

I am not entirely sure you are understanding what I am doing... The law requires I (my company) archive and retain these files, as they are medical documents. There will be 80TB of real data stored on it over the next 18 months or so. My options are to purchase an enterprise SAN solution, which on the cheap side for this capacity will run over $200K, or I build something custom for significantly less. The extra processor and compatible board added a total of $600 to the total cost of the build, and I get a significantly better motherboard for the server. If this was a government agency, they would have contracted with EMC or Dell/Compellant and dropped 2-3 million dollars on a tiered SAN system.

[rooster]

I am not entirely sure you are understanding what I am doing... The law requires I (my company) archive and retain these files, as they are medical documents. There will be 80TB of real data stored on it over the next 18 months or so. My options are to purchase an enterprise SAN solution, which on the cheap side for this capacity will run over $200K, or I build something custom for significantly less. The extra processor and compatible board added a total of $600 to the total cost of the build, and I get a significantly better motherboard for the server. If this was a government agency, they would have contracted with EMC or Dell/Compellant and dropped 2-3 million dollars on a tiered SAN system.

So weird to see someone talking about my related job.

We sell maintenance for EMC/Isilon HW and SW. And yes, we deal with a lot of government accounts around the world.

[Vindictus]

muh tax dollars

[Lord Dave]

I bet government agencies do shit like this all the time, wasting money on overblown systems just for shits and giggles. What a jackass move.

I am not entirely sure you are understanding what I am doing... The law requires I (my company) archive and retain these files, as they are medical documents. There will be 80TB of real data stored on it over the next 18 months or so. My options are to purchase an enterprise SAN solution, which on the cheap side for this capacity will run over $200K, or I build something custom for significantly less. The extra processor and compatible board added a total of $600 to the total cost of the build, and I get a significantly better motherboard for the server. If this was a government agency, they would have contracted with EMC or Dell/Compellant and dropped 2-3 million dollars on a tiered SAN system.

As someone who works for a government agency (albeit a school district) I can assure you that we would do no such thing as we don't have the funds.
Our e-mail is archived via google.
Our student records are archived via both paper and eSchool Data, our student database program (cloud based).

So yeah, we went the cheap solution.

[juner]

I bet government agencies do shit like this all the time, wasting money on overblown systems just for shits and giggles. What a jackass move.

I am not entirely sure you are understanding what I am doing... The law requires I (my company) archive and retain these files, as they are medical documents. There will be 80TB of real data stored on it over the next 18 months or so. My options are to purchase an enterprise SAN solution, which on the cheap side for this capacity will run over $200K, or I build something custom for significantly less. The extra processor and compatible board added a total of $600 to the total cost of the build, and I get a significantly better motherboard for the server. If this was a government agency, they would have contracted with EMC or Dell/Compellant and dropped 2-3 million dollars on a tiered SAN system.

As someone who works for a government agency (albeit a school district) I can assure you that we would do no such thing as we don't have the funds.
Our e-mail is archived via google.
Our student records are archived via both paper and eSchool Data, our student database program (cloud based).

So yeah, we went the cheap solution.

Yeah, I was mostly thinking of the Federal Government, specifically DoD. I know school districts and local governments don't have that kind of scratch.

[Lord Dave]

I bet government agencies do shit like this all the time, wasting money on overblown systems just for shits and giggles. What a jackass move.

I am not entirely sure you are understanding what I am doing... The law requires I (my company) archive and retain these files, as they are medical documents. There will be 80TB of real data stored on it over the next 18 months or so. My options are to purchase an enterprise SAN solution, which on the cheap side for this capacity will run over $200K, or I build something custom for significantly less. The extra processor and compatible board added a total of $600 to the total cost of the build, and I get a significantly better motherboard for the server. If this was a government agency, they would have contracted with EMC or Dell/Compellant and dropped 2-3 million dollars on a tiered SAN system.

As someone who works for a government agency (albeit a school district) I can assure you that we would do no such thing as we don't have the funds.
Our e-mail is archived via google.
Our student records are archived via both paper and eSchool Data, our student database program (cloud based).

So yeah, we went the cheap solution.

Yeah, I was mostly thinking of the Federal Government, specifically DoD. I know school districts and local governments don't have that kind of scratch.

Oh yeah.  Except the IRS.
Worse part: The companies who contract will over charge....

[xasop]

If the data is really that important, you probably want to build two storage servers, one to keep backups (unless the data is already stored somewhere else, and will continue to be available there). RAID won't protect you against filesystem corruption or accidental deletion, nor more severe problems like data centre floods (although unless you have two data centres, that's going to be difficult to mitigate).

You also have the option of archiving the data to tape, or even encrypting and archiving it to Amazon Glacier, just in case all else fails. I'm guessing you've considered these issues, but they're worth mentioning regardless.

[Rushy]

If it is only for storage, why does it need two processors?

It doesn't.

two Xeon E5's

I R Confucius

[juner]

If the data is really that important, you probably want to build two storage servers, one to keep backups (unless the data is already stored somewhere else, and will continue to be available there). RAID won't protect you against filesystem corruption or accidental deletion, nor more severe problems like data centre floods (although unless you have two data centres, that's going to be difficult to mitigate).

You also have the option of archiving the data to tape, or even encrypting and archiving it to Amazon Glacier, just in case all else fails. I'm guessing you've considered these issues, but they're worth mentioning regardless.

Yes, I have covered all of the scenarios. The data going on this server isn't actually owned by us, it is imported from other hospitals who retain their own copies. It is more for convenience for our doctors, as it is faster to retrieve locally than from a hosted archive I have that retains archived data owned by us. It is also incredibly cheaper to not use hosted as our hosted archive is well over 100TB right now. Technically, I don't have to retain this data for 10 years, but we do it as a favor to our hospital partners because their systems can be sketchy sometimes. The data does have to be retained, I am just not on the hook if it gets lost for any reason.

A few months back, I built similar servers for our production data that I do need redundant copies of. Two servers with 6x6 RAID sets were built as the data demands weren't as large. There is one stored at each of my data centers which are connected by dual 10Gb trunks. Replication software is installed that acts as a load balancer as well, so all nodes point to a single path, but that data gets simultaneously written to both servers. The replication software handles any changes so everything is in real time and either server could go hard down and it wouldn't affect anything. I migrated about 12TB of data from very expensive SAN storage, which really wasn't designed for the job, but the custom built servers are running very well.

I R Confucius

It doesn't need two, but it has two.

[Rushy]

It doesn't need two, but it has two.