Not logged in » Login
Oct 26 2020

A little insight into the world of storage arrays and a few things to consider when choosing one over the other

Image

One of the key benefits of using a storage array is consolidation of storage resources at a centralized location for greater effectiveness. On a generic level, these resources could be drive types, ports, controllers (single/dual), and enclosures. At the next level, these resources could include software features which are supported on the respective system (either default or licensed versions).

That said, however, there are options available in terms of which storage array to choose – traditional storage, software defined storage, hyperconverged storage etc. and within each of these options there are many variations to look at. As a customer one could be spoiled for choice, but it makes life simpler to evaluate these storage options and then make an informed decision on which of these makes most sense in the data center environment, in question.

For the purpose of this write-up we will look at traditional storage and the variations available, as well as why some of these variations, even though ‘traditional’ are still in demand. Then we will look at the considerations which might come in handy when choosing one variation over another.

 

All-hard-drive storage array

It might be surprising for few to see a mention of all-hard-drive storage arrays in this write-up but, as of the end of 2019, a little over 5% of the overall WW storage revenue (as per IDC) comes from all-hard-drive storage arrays. This makes sense as these arrays might find a good use case with customers who are looking at an initial centralized storage array, or a storage array on secondary sites (where configuration replication of primary site is not a mandate), or an exchange server, or a web server (where number of users are limited).

 

Hybrid storage array

As the name suggests these are the arrays that can house SSDs and SAS/NL-SAS drives. Different drive types can be clubbed in different enclosures or, with some newer arrays, they can be mixed within the same enclosure (as well), as long as we are looking at the same form factor.

These arrays makes most sense for someone looking for performance and capacity from the same storage array. This can be achieved by having a tier of SSDs, which can deliver the performance one is looking for, alongside a tier of SAS or NL-SAS drives which can fulfil the capacity needs, thus making these arrays a perfect budgetary solution all-in-all.

However, there is more to hybrid arrays than that, especially when it comes to a few other software features.

Depending on individual workload or application requirements, options are available to have different drive types tiered to cater to different requirements either periodically or manually (in preset manner). This feature, also referred to as tiering, is available in two to three tiers in arrays – entry to mid-range, and with enterprise class arrays there are options available which can offer up to four tier levels. It could be a default or licensed feature, depending on the offering from respective vendor.

 

 Image

 

As a next level, arrays can also map the applications demanding maximum number of IOPs (and minimal latency) to the fastest drive types/tier available, somewhat less demanding workloads to the next drive type/tier and the least demanding workloads to the next drive type/tier. This feature, otherwise known as Quality of Service is generally available from newer generation entry arrays onwards. Although this feature can be handled manually, it makes most sense when it is automated. It could be a default or licensed feature.

 Image

All-flash storage array

Again, as the name suggests, these are the arrays built to host all SSD drives.  

These arrays makes most sense for someone looking for performance – in terms of maximum IOPs (inputs/outputs per second) and minimum latency. Given how SSDs work and the fact that the price of SSDs has gone down in recent years, these arrays are preferred by organizations who are considering a good return on investment with an understanding that speed and timely action are extremely important for their customers. Example could be a customer accessing a banking app to make online payments.

As to how they give an excellent return on investment over the years, let’s say an organization needs a system that can deliver 200,000 IOPS to run an application; with an all-flash array this would be possible with only 10 high-end SSDs compared to 1,000 HDDs on other array types. Not only will this result in direct savings from reducing the number of SSDs, it will also result in substantial savings in power and cooling, management etc. which ultimately will have a big impact for  service providers, customers with server farms, etc.

Having said that, there are two criteria’s in particular that you might want to capitalize on in order to derive the best value from your all-flash array.

Firstly, the ability of a storage array to identify the data which has been written twice so that, instead of a repeat data write, only incremental data actually gets written (with pointers to original data set). This feature is known as deduplication and it results in saving further space on a per-drive basis. We do have arrays available today in which this feature is built as in-line, and with an option to activate or deactivate it depending on the application’s requirements. Generally default, it could be a licensed feature as well, depending on the offering type from the vendor.

Secondly, the ability of a storage array to further pack data in a form and format which takes the minimum possible space on a drive. This feature is known as compression and it results in further capacity optimization by using less drive space. The difference between deduplication and compression is that, with compression system identifies a repeating data pattern, re-encodes it into something which conveys the same meaning but all without taking as much space. Again it could be a default or licensed feature, depending upon the offering.

 

Image 

 

NVMe storage array

These are the storage arrays which support the use of NVMe drives.

They make most sense for customers when latency is of paramount importance for business applications. And this is possible with the use of NVMe drives and how they process IO requests. Given the fact that NVMe maps I/O commands and responses to shared memory in a host computer over the PCIe interface, they can deliver up to 6x better performance when compared to classic SSD drives.

Below is the summary of offerings available today from different vendors, under the NVMe tab:

  • Earlier generation all-flash (or hybrid) arrays. They may have an option for NVMe SSDs along with SAS as of date. Generally targeted/positioned in the entry storage segment;
  • New generation all-flash (or hybrid) arrays. They partially have the feature availability (connectivity) of an NVMe storage array however SSD drives remain a default option;
  • Storage arrays built from scratch with front end and back end support for NVMe along with NVMe drives. Also, support for NVMe-OF is either available or will be made available in future.

Though NVMe arrays are the in-thing right now, in order to derive their true value NVMe protocol should be supported all the way in the data center, starting from compute-to-network-to-storage, otherwise bottlenecks at any stage might affect the expected performance. 

 

Next are few things to consider when making the decision to invest in one of the above storage arrays. While we all want the very best, it is important to understand what we actually need vs what we want, and with that in mind, here are few basic considerations:

 

  • How easy is the system to deploy along with data migration guidelines? (To consider in the case of migration from older to newer array.)
  • What are the requirements for managing the system; how easy it is to manage?
  • How effectively will it embed in the existing infrastructure in data center (considering virtualization environment and other applications)?
  • Will the new array help deliver on the current and future performance requirements (IOPs and latency)?
  • How much effective floor space (data center) will the new array take up and what will be the additional impact on power and cooling, if any?

 

We at Fujitsu understand that as a customer you will have different data center requirements and those will be based primarily not only on applications run in-house but also on the nature of your business. And in order to address each of these individual storage requirements we can provide an answer on which solutions can be custom made for you. To find out more about the data storage options available from Fujitsu, please visit this page.    

 

 

Megha Shukla

 

About the Author:

Megha Shukla
close share field

Megha is part of the Product Marketing team, Fujitsu Global business. She has several years of work experience with multiple MNC's across Enterprise IT space...

SHARE

Comments on this article

No comments yet.

Please Login to leave a comment.

X

Please login

Please log in with your Fujitsu Partner Account.

Login


» Forgot password

Register now

If you do not have a Fujitsu Partner Account, please register for a new account.

» Register now