We had a presentation of flashray and I was kinda left with a 'meh' impression and I think that Georgens is wrong with his 'you can't have idle data on flash' because with flash you may pay more per gb but you pay less in power, cooling and rack space.
NetApp's CEO Tom Georgens philosophised about flash storage in the company's latest earnings call with analysts. The top boss said non-volatile NAND memory isn't the be-all and end-all in storage: what matters is where active and inactive, hot and cold, data is stored, and how dear each of those holding pens are. “Flash, on a …
At last check, an 800GB SSD (3 watts) actually consumed more power, cooling and rack space per GB than a 4TB HDD (7.8 watts).
Man selling inferior product claims superior products aren't really superior
Can somebody please send him a link to the description of "Confirmation bias"?
Hmmm...he's not heard of RRAM then?
If customers are not buying that flash is cheaper than disk, why are Seagate and WD reporting zero growth in 15K SAS? When you look at a storage system to get the identical performance you get from flash you need far more drives and need to use RAID 5/6. The drives will likely be short stroked as well so there will be plenty of unusable capacity. Insofar as comparing flash to SATA that is comparing a Porsche to a Mack truck. They are designed for completely different purposes.
Speaking of NetApp, anyone heard if they are going to refresh the FAS line as EMC is doing with the VNX? Any idea of the new model numbers and what their relative performance is?
NetApp MIdrange Filer Refresh
>>Speaking of NetApp, anyone heard if they are going to refresh the FAS line as EMC is doing with the VNX? Any idea of the new model numbers and what their relative performance is?
Given that Netapp only refreshed the 32XX line with 3220 and 3250 less than 12 months ago. (November 2012) And new 62XX models FAS/V6220, FAS6250 and FAS6290 This Year. (Febuary 2013)
I doubt there is any new midrange Filers coming from Netapp for a while.
15k SAS is flat, and single level flash probably isn't doing much better
The time frame the NetApp guy was looking at was probably 3 years or so, and during that time dumping most data to big, very cheap, slow disk, then cleverly caching it in RAM and Flash in controllers, looks like a reasonable approach for many applications. It can provide superior performance to plain flash (RAM is so much faster than current flash) while being cheaper.
That 15k SAS is flat, and single level flash probably isn't doing much better makes sense.
What's growing are the denser, less expensive versions of storage technologies. It seems like systems that once collected a binary value using a single bit of storage are now archiving a Yes/No question as a 1920x1080 video of someone leaning into their webcam and saying "No" - which then has to get replicated to D2D backed storage in two or more clouds (once your cloud provider goes chapter 11 and sells off the boxes holding your data once, you learn to replicate). Even after compression and de-dup that bit has become a couple meg scattered across 4 sets of blocks!
Granted, that's a reductio ad absurdum worst case, but video and imagery (and backing up do disk) have been taking away a lot of what increasing density and falling costs have been giving to storage. It's why the NetApp guy was saying that they see demand very fast cheap storage growing faster than slightly faster expensive storage.
Why are Seagate and WD reporting zero growth in 15K SAS?
From what I am seeing as a storage sales rep, customers are largely moving to 10K 6Gb 2U Shelf SAS drives down from the 15K 4U SAS drives because they take up less rack space, use less power and offer more bandwidth than the current 15K drives. I cannot remember the last time I sold a new system with 15K 3 1/2" SAS drives. The only time I see customer purchasing 15K SAS is to incorporate them into an existing 15K SAS array. That is why you are seeing 15K SAS sales as flat or declining. Not because they are buying flash, but because they are moving to 10K 6Gb SAS drives. Also the 15K 3 1/2 3Gb SAS drives will eventually be going away, so customers buying new systems are avoiding them to keep from painting themselves into a corner.
"If customers are not buying that flash is cheaper than disk, why are Seagate and WD reporting zero growth in 15K SAS?"
Because sales all the way from top end to low has been poor of late, and SAS is a mature market (so the initially high growth of having 0 on the market one month, more the next month, and more the next is done).
If one needs very fast performance, flash is a good way to go. Some people seem desperate to make flash cost-competitive with disks though, and the fact of the matter is it's simply not.
No law that states spinny things will always be cheaper
And should it come to pass that a mass-market cheap high capacity high performance solution emerges which consumes no power except when reading or writing, then it's disky-bye-byes.
I hope it's not too long now!
Is this because NetApp isn't in this market?
Fear, Uncertainty and Doubt may hold back the tide for a little while, but it is interesting that a CEO feels he has to step up and be economical with the truth. Reading between the lines I guess he's making a play for the slow cheap market since the high performance market is passing him by.
I read a couple of VMware reports on how they made 1 Million IOPS benchmarks, the first used 8 servers and 680 15k rotating disks. The second was eight months later and used one HP DL380 running vSphere 5.1 and two Violin 6616 arrays. That's a million IOPS from 9U of rack space. More recently they ran 800k IOPS with one 6616.
The biggest 3.5" drives I see commonly for sale are 4TB and I can get a 4U rack to hold 20. So I can get 80TB in 4U or 20 TB per recking unit. The latest Violin array they just announced is 64TB in 3U so the density is already ahead. Of speed is your need then the VMware benchmark would suggest you need 136U to get enough disks to run 1M IOPS and you'd still have comparatively awful latency, ms instead of us.
So in some ways I guess NetApp has a point, currently mass storage is probably cheaper to buy in disk. But a quick look at the history books would show how mechanical development has gone compared to electronic development. Would you like to buy a tape drive? So what high end business is going to be using electro mechanical disks in 5 years? Even this cheap home PC runs from a SSD now, I was going to replace the machine until I replaced the disk and found I didn't need to any more.
Nimbus says Hi. "285TB in 2U" (although I'll have a coke with that salt).
I agree with him to a large extent, there will be a place for each technology and the key will be having the IP to know where to put it and move it around.