There’s been some chat going on about disk redundancy in Seedboxes. So how does the landscape look like?
Google search of RAID5 Seedbox yields pretty much no results other than our services! Some results for RAID10 seedbox query, but all those offers are either years old, or really not cost-effective at all, same goes for Raid Seedbox search.
Why are there no RAID5/10 Seedboxes?
Cost, Cost and Cost. The bulk of Seedboxers request for maximal value of the service, and most don’t consider that data important enough for redundancy.
Yet, there are users who assume by default that RAID or backups should be included in all hosting services – despite company in question does not promise any data redundancy, usually quite to the contrary.
When users are asked, almost no one is willing to pay even a dime for redundancy. Infact, some users even have demanded RAID0 config even if there is no cost difference!
It all boils down to economics for the RAID Seedbox
It’s all down to economics really, most Seedbox providers rent their hardware from other datacenters. Yes, we do this a lot as well, but that’s only because we can only grow so fast our own datacenter operations.
So the common low end offers are for 2x Disks or single disk, which leaves you only option only to either have RAID10 or backup to another server. Not very cost effective. The price for 4 disk system increases vastly, often times 4x2TB server costs 3 times as much as 2x2TB server! This is unbelievable – but that is how the market is!
Even assuming you are lucky one and can get a 4 disk system for the price equivalent of 2 disk system, and can do RAID5, seemingly you are “wasting” 25% of the storage space, which means end user price is at least 35% higher than RAID0 offering. If you could offer 1TB for 20€, it would go down to 750GB for 20€.
Since there is pretty much standard pricing which is about 14€ for 1TB now, let’s scale this up!
Simplifying, your servers with 4x3TB cost 84€, and you have 150 users with revenue of 2100€. RAID0 requires 13 servers, RAID5 requires 17 servers, RAID10 requires 25 servers. RAID0 servers cost 1092€ leaving you a profit margin of 1008€ for all the overheads (staff, insurances, accounting etc etc.!), RAID5 servers would cost 1428€ leaving you a profit margin of 672€.
Doing some maths reveal the RAID0 has profit margin of: 52% and RAID5 has profit margin of: 32%. That, is a huge difference!
When you own the hardware
Things change vastly, your fixed costs per server are electricity and space. Bandwidth usage is per user.
In our case each 4 disk server costs in electricity roughly 11€, and housing costs about 20€ a month for a total of 33€ a month. If there is RAID0 data is lost on all disk failures, which causes an extra 1-2 hours of work, or the need to have significant number of spare servers always. Having spare servers is not really an option, so everytime a disk fails someone needs to go on-site promptly, and the cost of this inflexibility is far greater than having RAID5.
With the same 150 users example and using 3TB drives, cost for RAID0 is: 429€, cost for RAID5 is: 561€ in fixed costs. Certainly, disks and nodes do cost, but we got to remember we use those same disks for 3 years. Our per server costs are: Fixed (electricity + housing) plus hardware plus installation & maintenance costs. With RAID5 we lower the installation & maintenance costs significantly. With 13 servers in the example we can expect 4.65 disk failures annually – if there is no issues causing spikes in disk failures.
The cost of all RAID0 node disk failures is a lot higher than with RAID5. With RAID5, we go on-site whenever it suits our schedule and got other work to do onsite, users get to keep using their service in the meantime, and when the disk is swapped it usually only means 5-6 hours of rTorrent downtime only.
RAID0, we have to jump to the site whenever it happens everytime it happens, being it’s 30mins drive per direction alone, and OS installation + disk diagnosis easily takes 2 hours, and all customers on that server loose the data we can expect 1 or 2 will cancel due to this. AND, we do not get to stress test & verify disk quality prior to provisioning the new disk!
So, 3hours of working time + loose of revenue. If that happens to be 3AM Sunday, those 3 hours easily cost 230€ alone! The lost revenue costs minimum 100€ +, so this is 330€ for a disk replacement! Never mind lost night sleep. The minimum cost of instance like this is close to 200€, so we can average it to something like 250€.
When this happens 5 times a year, the annual costs is 1 250€ for just 13 servers! Imagine, if you have 130 servers to maintain! You’d have 46 full node data losses per year! You’d have all nodes with complete data loss every 2½ years roughly.
Annual cost for RAID5 in fixed costs for the 13 servers is 1 584€. Bottomline, costs wise it’s not such a big difference!
Finally, the cost for 13 servers is only 27.83€ per month or 2.14€ per server. Of course, this does not account for the hardware costs or actual harm done to reputation & revenue over the long term, 100€ lost revenue is very very conservative, considering the lost users in a instance like this are usually long term users, some of which stick with the same company basicly forever they need the service. Imagine loosing single 5 year user: Even with our cheapest entry level plan at 5€ a month that is 300€ ! Then account for the bad reputation, single disgruntled user on average makes sure that several potential customers do not signup, research shows that in some industries this can be as many as 7 ! Ultimately, it costs us less to offer a better service than the market standard is!
Caveat emptor, we have optimized everything at all levels for this type of usage. Many seedbox vendors have to use individual drives for resource use control. Our methods would not work as efficiently on say a 12 drive system and we do prefer smaller unit size.
Also note that not all of our servers are our own, we still have a vast quantity of RAID0 seedboxes as well where it simply is not economical to offer RAID5/RAID10, but Dragon series is guaranteed to get either RAID5 or RAID10. A lot of servers for Super100 are RAID5, but almost none for the Super20 entry level plan, simply due to server size getting too large for the price point and to ensure high performance at all times by not overloading the servers with too many users.