STORAGE

  • 07/09/2015
    12:30 PM
  • Rating: 
    0 votes
    +
    Vote up!
    -
    Vote down!

Selecting Storage: Start With Requirements

There are lots of buzzwords on the storage scene these days. But don't even think about buying in until you truly understand your business' storage needs.

A common question I am asked is, "What is the best storage technology?" My routine answer is, "It depends!" During my recent Interop Las Vegas session "Smart Shopping for Your Storage Strategy" I addressed this very question. 

I often observe people deciding on the best storage before the questions of requirements, needs and wants are even mentioned. This is usually tied to a game of Buzzword Bingo -- you know, that game where somebody starts naming off industry trends and buzzwords until you say, "Okay, that's enough; let's focus on the challenge or opportunity, and then come back to the buzzwords." Buzzword examples include object storage, software defined storage, hybrid, , flash SSD, AFA (all flash array), APA (all purpose array), converged or hyperconverged infrastructure, scale-out, scale-up, VSAN, elastic, ephemeral, data lake, pond or ocean…. You get my point.

Thus the best storage solution will be the one that protects, preserves and serves your information for your organization, as well as one that works for you -- instead of you working for it. By that I mean that the best storage solution, product or technology should adapt to and fit your environment. This means being able to meet your various application performance, availability, capacity and economic (e.g. PACE) needs, while also complementing existing technologies, tools and expertise.

Figure 2:

The best storage depends on what you need the storage to do. Ask yourself the following questions:

  • Will the storage solution be used for all of your storage needs, or is it for a particular business function or application?
  • Do you need to converge your servers and storage into a consolidated platform, or do you need storage that can be shared and accessed from many different virtual and physical servers?
  • Does your environment require high performance (bandwidth, IOPs or low latency), or high capacity for static data?
  • What type of availability and durability does the storage need?
  • What associated data services (snapshot, replication, thin-provisioning, access, etc.) and management capabilities do you need?

The best storage, product, service, or technology and how you package it will depend on what you need (or want) it to do. A solution for primary active storage will be different from one for secondary bulk storage or tertiary data protection. Storage specific to an application also has different requirements that storage provides a deterministic, predictable quality of service overall. Or perhaps you simply need the most storage space capacity at the lowest cost because you will be configuring it with other data and storage management software tools to improve performance, availability and management.

Figure 1:

During my Interop session on storage decision making, one of my major recommendations was to take a step back and outline your exact storage needs. Define your business' needs vs. wants (what you would like to have, or what somebody else wants or needs you to have). Once you develop solid requirements, you can start narrowing down the field of products. You may lean toward single storage solution, such as an APA or hybrid system that is capable of supporting HDD and SSD with various data services, flexible access and management tools. Or you may realize that point storage solutions are a better fit so that you can target specific applications.

There is more to storage decision making than calculating cost per gigabyte or terabyte. But while it may seem like storage is complex, you can apply the same decision making strategies used for other technologies and investments. The key is being an informed decision maker and understanding your needs along with the options. Then you can make an apples-to-apples comparison, rather than trying to compare apples to oranges.


Comments

storage shopping

Thanks for this practical guidance Greg. There really are a lot of buzzwords tossed around in the storage industry today, which could make storage shopping complicated and obscure what a company really needs.

Re: storage shopping

No worries Marcia, you are spot on that the industry has and likes to toss buzzwords around, particular new industry first and revolutionary ones ;)... Certainly all of the options can and does lead to confusion, and for chaos or confusion leads to oppourtunties while for others it results in frustration. A while back I was asked if Compute Data Storage is complex which it is or can be, if you are not informed as well as know your needs. Its not that different than shopping for an autombile, cell phone, laptop, or many other things that you may not be deeply familar with. Dont be scared of server storage I/O hardware, software or services and technology shopping, be preapred and be a smart shopper, hmmm, I think I know theme for follow-up piece ;)...

 

Re: storage shopping

I'm looking forward to your next post! 

Re: storage shopping

So am i... ;)

BuzzWords offer no solution for Real Storage Questions

Thanks Greg for addressing the deluge of buzz words going on in the industry. It really is getting to the point of being counter productive. As a tech professional, I am spending more time trying to understand the latest buzz word instead of working on core issues.

But I want to get your feedback on a real issue - that of storage and optimization of throughput. That is a real issue for many engineers and buzzwords offer no solution to this on-going problem. What do you feel is the best storage configuratio for this ?

Re: BuzzWords offer no solution for Real Storage Questions

@ClassC no worries, sorry for delay getting back to use, need to sort out an issue with notifications on my end.

U r correct the industry loves buzzwords which can cause chaos or confusion, however in chaos and confusion there can also be oppourtunity, hmm, perhaps that will be a future post theme? ;)

Ok, lets get to you note about optimizing for bandwidth or throughput, concur it is a real issue for many environments and applicaitons, and from my experience, not just classic high performance computing (HPC) or supercomputing. As data gets larger the average I/O size or IOP is also growing which means more bandwidth or thoughput focus (or concern). Many of the solutions are optimized or focused around high IOPs of small I/O size or latency including cache tools, hardware, and software.

You do have options depending on what your specfic applicaiton and bandwidth acceleration needs are, including cache or micro-tiering tools that work or can be tuned for larger I/O size and data transfers vs. smaller IOPs. For example, Enmotus has a micro-tiering tool that optimizes for reads and writes (you can tune for either, or both) that can also be tuned for small IOPs, or larger IOPs to improve on bandwidth. Likewise hardware and software based RAID, erasure code, dispersial and FEC (hey, that was a short buzzword banter ;) have the ability to tune for larger I/O size via parameters such as chunk/shard/stripe among others. Also you can configure your filesystems or software defined storage aka volume managers, virtualization layers, etc to optimize for bandwidth vs. IOPs or both.

As for the best storage configuation for bandwdith, it depends, tell me more about a given applicaiton can and provide various options, otherwise its just a list of vendors products, solutions or services that may or may not be applicable, however its fun and easy to toss names, terms and buzzwords around ;)...

Cheers gs

 

Fusion IO : Just Hype ?

I have seen many engineers rant about Fusion IO to improve speed and I have seen it employed but I didn't think the supposed performance increase was worth the cost.

 I actually thought it was overrated.   But this is the type of solution engineers are pushing.  What are your thoughts about this type of solution for increase performance ? 

Re: Fusion IO : Just Hype ?

@ClassC keep in mind that the best I/O is the one that you dont have to do, the second best is the one with the least impact and cost effective. Also keep in mind that a little bit of flash SSD in the right place can go a long way while be cost effective as a cache (or storage) vs. costing a lot of cash. The Fusionio (e.g. FIO) now SANdisk cards are good for some applicaitons, likewise the various second and third generation cards from other suppliers including Intel, Micron, Samsung, Seagate (via LSI/Sandforce), WD (via Verident) among others are also good for some things, while other solutions may be more applicable for others. Some of the cards are more expensive vs. others, some have had more marketing promotion or hype behind them, hence why you might see some people pushing them or other things more than those they are not aware of, or that do not have the same marketing focus behind them.

Most of the hype marketing has subsided in the PCIe AIC flash card space as they have become more mainstream, granted, watch for a new flurry round of activity as more vendors roll out their NVMe enabled solutions in the comming months. However there is some similar activity occuring in the all flash array (AFA) and other flash storage systems marketing game. How much you pay is also a function of how much you are are willing to pay for a given solution, something more popular may be more expensive if you feel the need to buy it, otoh, there are value price performers that may not be the top marketed brand, yet also can meet your needs. Your needs, requirements and compeition options are key and if you can stick to what you need vs. what you want, or what a vendor or solution provider wants you to need, you can increase your applicaitons performance in a cost effective way.