SolidFire all-flash-array anyone?

Thuleman

Supreme [H]ardness
Joined
Apr 13, 2004
Messages
5,833
Does anyone have one of these http://www.solidfire.com/storage-system/ in production?

Pricing seems attractive, it's a startup but not that new. They have been shipping units for a couple of years if I read that right. Fairly negligible market share, but the tech looks impressive.

I like the 90 second QoS explanations here: http://www.solidfire.com/storage-system/qos-benchmark/ specifically that capacity can be separated from performance.

It all looks good, but the proof is in deployments of which there aren't that many. I seem to have read somewhere that they had $4M in sales in 2013, probably(?) more in 2014. That's still just a drop in the bucket.
 
Good gear. The QoS is really targeted toward service providers but can be used by anyone. I know several people over there. We don't sell it since we don't want to pick up another AFA but I would give it a look. They were slower starting because their products were mainly for SPs. You had to put in a large initial footprint. Now you can start much smaller which makes it a lot more attractive.
 
Good product, expensive, overkill for most customers - local company to me :). If you like that QoS, take a look at what Coho and Tintri are doing as well.
 
Good product, expensive, overkill for most customers - local company to me :). If you like that QoS, take a look at what Coho and Tintri are doing as well.

Is it really that expensive? I seem to recall that 4-node 36 TB raw setup for less than 150k. In comparison that appears to be inexpensive for AFA even if you can't actually use all of the 36 TB.
 
Add in raid, node redundancy, etc - it gets up there pretty fast, for the amount of space you get out of it.
 
Is it really that expensive? I seem to recall that 4-node 36 TB raw setup for less than 150k. In comparison that appears to be inexpensive for AFA even if you can't actually use all of the 36 TB.

That is effective capacity. I got quoted for the base model, SF2405, in a 5 node cluster. I was told it yields a 43 TB effective capacity, but if you look at the specs, each node is 2.4 TB raw. That comes to 12 TB raw across all 5 nodes and the price is just about $200k.

Needless to say I went with Tintri and couldn't be happier.
 
They are probably telling you the effective capacity after dedupe and compression. You don't know the exact amount until you get some of your data on it. If you're worried do a PoC or have them give you a written guarantee on a minimum dedupe/compression ratio.
 
Needless to say I went with Tintri and couldn't be happier.

I get that you are happy with Tintri, but Tintri couldn't be any further from a SolidFire in both the AFA as well as the architecture realm. They are literally only comparable on the basis that both store data.
 
I get that you are happy with Tintri, but Tintri couldn't be any further from a SolidFire in both the AFA as well as the architecture realm. They are literally only comparable on the basis that both store data.

Thanks. I didn't say you had to go with Tintri. I was pointing out the capacity and cost as I also looked into SolidFire.
 
I briefly looked at it. It looks like a decent product, but I see it as a product looking to be bought out by the main storage providers and really not in it for the long haul.

re dedup and compression. Don't believe what ever the sales guys are saying. get a POC with your data. Pure and EMC are telling me with their arrays I'll see at least 5 to 1 compression. I don't buy it because im lucky to see 3:1 compression via my backups on data domain.

every ones data is different.
 
We have a 8 Node 4805 cluster and so far very impressed with the QoS and the sustained IOPS for different work loads. In POC mode now, and should be getting 2x in compression, dedupe and thin, but not close.

The bad:

The 4-to-1 is completely not possible with a mixed workload. Even with our 400 VM's on it, where 180 are clones from a known template. Were still seeing way less then their sales is touting. VM's consist of 4 Exchange servers, approx 4TB of mail, a few small SQL servers with less than 100G databases, bunch of operations VMs, 4 file servers with over 5T of files, mostly documents and some media. The remainder is 50/50 Citrix of XenApp/XenDesktop VM's for clients.

8 Nodes of 4805 should yeild over 120T but were only at 47T used and over 64% full on block storage. Not going to be able to stuff all in what we wanted, no looking good for us. Its the most expensive all flash and not able to achieve anywhere close to what sales states.:mad:

SF2.jpg

SF1.jpg
 
Last edited:
Yeah..see...I *HATE* it when companies include thin provisioning in to the logical capacity. Pure does this too. We don't when discussing XtremIO so I often have to go talk to customers and show them why Pure is supposedly getting much higher.
 
Im doing a POC this second with XtremIO vs Pure.

with my data on XtremIO I'm seeing 1.8:1 and no compression when EMC said at least 4:1, Pure is telling me 5:1. My pure POC is next week. The POC is really whats going to be the make or break for either of them.

I think folks are hyping the storage savings a little to early right now. I think dedupe, and compression is awesome and all.. but its not what the sales guys are claiming.
 
dedupe and compression are fine if you are running a lot of redundant VM's. ie: virtual desktops, term servers and such. If you are hosting databases with large working sets then it's pretty much meaningless.
 
dedupe and compression are fine if you are running a lot of redundant VM's. ie: virtual desktops, term servers and such. If you are hosting databases with large working sets then it's pretty much meaningless.

That's really what gets me. Fine, VDI is a decent candidate for flash. After that it's database workloads, yet those simply don't hit the advertised dedupe rates. This isn't even an argument of vendor A dedups better than vendor B, they all roughly equally don't dedup. All the conversations around dedup end up being disappointing by vendors overstating their solution's capabilities.

I'll probably bring Solidfire in for a POC in September. Really like the "Isilon-like" architecture.
 
That's really what gets me. Fine, VDI is a decent candidate for flash. After that it's database workloads, yet those simply don't hit the advertised dedupe rates. This isn't even an argument of vendor A dedups better than vendor B, they all roughly equally don't dedup. All the conversations around dedup end up being disappointing by vendors overstating their solution's capabilities.

I'll probably bring Solidfire in for a POC in September. Really like the "Isilon-like" architecture.

Which is why we went Tintri for our database storage. The performance for our workload is great for the amount of space it offers at a price that's cheaper than an ADA of equal performance. AFA is just too expensive at this point in time.
 
Back
Top