What is Linode's view on providing SSD based VPS?

Will it happen, not happen? If there are plans, when?

SSD drives aren't big enough for a Linode yet and silly expensive, the biggest a quick search found was over $1000 for a 480GB.

_________________
Paid support
How to ask for help
1. Give details of your problem
2. Post any errors
3. Post relevant logs.
4. Don't hide details i.e. your domain, it just makes things harder
5. Be polite or you'll be eaten by a grue

It's a fair discussion to have I guess in 2011.

I believe there is a market for VPS with fast disk IO with lower storage capacity (e.g. Starter Web Apps). They maybe willing to pay a higher rental for such a machine. Or pay for a slice of a SSD?

I think this will be the year that more customers will start demanding SSDs in the datacenter.

I know my web app would run faster with pre-existing optimization on an SSD. Something that i might consider paying extra for (in 2011).

Really? Because, by far, the biggest complaint people post here is "want more disk", not "want faster disk".

Agree completely. We don't need faster disk performance, it's plenty fast as-is running 15K RPM SAS drives in RAID10. What we need is slower bigger storage as an option.

We want to have a file dump for our design department, and our linodes just don't have enough space to be practical.

Linode handily thrashes their competition in terms of performance and I've not had issues with or read of a lot of people having issues with disk I/O. I've seen a lot more people want more space.

There is absolutely no need to use expensive SSD.

Give me more storage, my speed is great.

Your best bang-per-buck on VPSen, at this juncture, would probably be to throw more RAM at your situation. You probably aren't going to see server-local RAID'd SSD storage in any reasonable quantities any time soon.

I'd bet my bippy there's already someone out there marketing SAN-based SSD storage with prices less than $1/GB/mo, but that kinda defeats the purpose.

_________________

/* TODO: need to add signature to posts */

If SSD performance is so vital for your workload then you'd probably be better off with a dedicated server anyway. Splitting up an SSD between virtual machine users just doesn't feel right at the moment - if you want blazing fast performance then sharing an SSD drive with other users just isn't going to cut it.

Even the lowest-end consumer SSDs haven't hit under $1/GB yet (~$1.50/GB is about as low as they get), so how would somebody make an enterprise-grade SAN for under $1/GB? I doubt they could even do it profitably if they were buying NAND flash themselves directly, and it would have to be MLC at those prices.

Note the time unit there: $1/GB/mo will cover that $1.50/GB in about six weeks. Less if it's oversold, compressed, or built atop lies.

_________________

/* TODO: need to add signature to posts */

/me waits expectantly for GoDaddy Super Mega Sexy SSD Shared Hosting

As a consumer, my experience with SSDs has been less than stellar. One of my SSDs failed 3 times within 18 months, requiring a total wipe to recover each time. I've got platter drives that are approaching 15 years old that still get the job done. When they work, though, SSDs sure are fun for a desktop machine.

But for a production server environment...personally I would avoid SSDs for awhile. I can't imagine trying to run my business with that kind of hardware uncertainty. I always wondered if a RAID 1 would provide some protection against disk failure, but that doubles the cost and I've never been sure that the failures (which seemed to relate to large data transfers) wouldn't hit both RAID disks at once - in other words, is it an electrical issue with the individual disk, or something in the firmware that could choke and kill both disks. SSDs from different manufacturers or with different controllers might alleviate that, but then you're stepping away from plug & play hardware in your datacenter.

I would sooner work on optimizing my code than throwing faster and potentially questionable hardware at it, but I know that's not always the simplest answer.

LOL@GoDaddy super mega sexy SSD hosting...

SSD have a finite number of r/w cycles that is much less than that of an HD. In the harsh environment of a vps, it would be unacceptable. While I'm sure it is off on the long side, it has been predicted that SSDs won't truly be competitive until around 2020 based on the expected performance, storage size and reliability increases in solid state and physical mediums.

I'm sure Linode management will stay current and offer the best balance of performance, reliability, and price based on my past two years experience with them.

Jeff

SSDs don't have a limited number of read or write cycles. They have a limited number of erase cycles. Yes, that tends to happen when you do a write, but not a read. Even a 64GB SLC SSD can handle "writing" ~8 petabytes before giving up the ghost (ignoring spare area, which bumps that a bunch higher). If you completely re-wrote the entire SSD every day in 512KB chunks, it should take you ~329 years to wear it out. Yes, that's a best-case scenario, as there are usage patterns that are less optimal, there's write amplification to take into account, and you might be writing more than the capacity of the drive every day. But I find it improbable that you could wear out an enterprise-grade SSD in even a busy server during the lifetime of the server that SSD is in.