Enterprise SSD options

The enterprise SSD comes in a variety of formats, such as within storage arrays and in servers via PCIe cards. Gartner's Valdis Filks talks about which SSDs go where in the storage stack.

The enterprise SSD comes in a range of formats, from those incorporated into storage arrays via dedicated solid-state drive arrays through to PCI card SSD for servers. They all provide huge performance increases over spinning HDDs but at a much higher cost

In this interview, SearchStorage.co.UK Bureau Chief Antony Adshead speaks with Valdis Filks, research director for storage technologies and strategies with research company Gartner, about the different types of enterprise SSD available, where they fit in the storage architecture and whether the enterprise SSD will supplant spinning hard disk drives (HDDs).

Play now:

Download for later:

Download the podcast with Valdis Filks
• Internet Explorer: Right Click > Save Target As
• Firefox: Right Click > Save Link As

SearchStorage.co.UK: What are the key types of SSD available to storage managers today?

Filks: The two types of SSD from a technological point of view are SLC and MLC NAND memory. That really isn't that important; [rather], it's how people package it inside storage arrays or if they sell you specific cards to put in servers.

Most vendors have now put SSDs in storage arrays. Initially, they were separate pools [that] you had to manage manually, but since about October [2010] every major storage vendor has been able to implement automated storage tiering and various policy-based systems where the storage array -- depending on the directions, policies and rules written by the storage administrator -- will move data in and out of the SSDs according to the performance requirements.

See our full enterprise SSD implementation guide

MLC vs SLC: Which flash SSD is right for you?


PCIe-based solid-state storage sees uptick in shipments and usage 


Solid-state drives vs hard drives: How to justify the cost of an SSD

Council adds SSD tier to Compellent SAN for new HillingdonFirst residents' card system 

There are other types of system where you can buy SSDs on PCI cards and stick them into one server. That can then only be used by that server and not many, many different systems [as it would be] in a storage array as a separate SSD pool.

There are also the traditional SSD vendors, who sell pure SSD boxes or SSD arrays. I call them JBOSes -- just a bunch of SSDs. Those can be shared by other systems, but they don't have any automated tiering; they are pure SSD JBOSes.

SearchStorage.co.UK: Where in the storage stack do the different types of SSD fit for the enterprise user?

Filks: They are what we'd call Tier 1. I can't really call it Tier 0; there are no real definitions for this. Different vendors make up different marketing terms. [SSDs] obviously fit below the controller cache in the storage arrays and above the 15,000-rpm Fibre Channel or SAS-type disks.

They generally cost around 10 times more than the fastest disk, but they're 100 times faster. And because of the cost difference, which is still quite considerable, they do sit above the stack of the [fastest 15,000-rpm] disks.

Some vendors are using SSDs inside the controller for a type of cache system, and the lines are getting blurred between the controller cache, which still is often DRAM or normal memory -- which is still even more expensive than the SSDs -- the SSD pools [and] the fast disks.

This is a good thing and a bad thing; we have so much choice now between which tiers or which parts of the storage hierarchy we can put data [that] we are probably not sure what we need to do, and in the old days we would do these things manually.

Nowadays, so many applications share these systems and the performance changes by the minute, there is only really one option to manage these and to try and work out where they should fit in the hierarchy. [That is] to have automated tiering or some form of automated caching system that moves the right data on the right [type of disk in terms of performance and cost].

SearchStorage.co.UK: Do you see SSD totally replacing spinning disk, and, if so, within what time frame?

Filks: I see SSD replacing HDDs in the time frame that I will probably retire, which will probably be in about 20 years.

So for all practical uses, [I think] -- and Gartner's position is -- that SSDs in the medium and probably the long term -- [say], five or 10 years out -- will not replace HDDs. That is in the enterprise. We will definitely see a lot more SSDs in consumer devices [such as] laptops and we've obviously seen them in mobile computing devices. But in the enterprise and in home [computing] devices, where we have to store large quantities, we don't see SSDs replacing HDDs for 10 years or more.

The problem here is that [with] HDDs, the price of capacity is dropping every year, as is SSD, so [in] our extrapolations and analysis, we don't see a point where SSD becomes cheaper than HDDs and manages to give us more performance for the same price.

[SSDs] will become more ubiquitous; we will see them everywhere, but they won't kill off the HDD as we know it today and for a considerable time in the future.

Read more on SAN, NAS, solid state, RAID