HP LeftHand capacity - Around the Storage Block Blog -
HP LeftHand capacity

By Jasen Baker, Storage Architect

I once heard that in communicating your opinion or differences, you should not use personal phrases, such as "you always", "you never", "every time", etc. These make broad, sweeping assumptions which seldom reflect the truth, especially when communicating differences in products. Attention to detail, such as quoting someone's name when providing a source of argument, or consolidating many options into a single unified calculation over simplify and often times mislead readers who are looking for educated facts instead of uninformed guesses.

As an example, take a look at online value calculators. There are calculators for mortgages, calculators for the national debt, ROI calculators, and even storage capacity calculators. They do their best to point you in a certain direction, to narrow down the scope of what you will be working with, but don't truly take in all the factors, hence why the infamous asterisk * exists!

In responding to a recent blog post referencing "LeftHand Capacity Calculator" crafted to demonstrate useable percentages of available capacity, it's supposed to ALWAYS, come out this way.  

To quote the blog, "That's because, regardless of how small or how large your LHN SAN, it's always:"

  Image "courtesy" of NetApp

There's that word again, always.

In storage, there are useable capacities that always occur. That always, is the space you lose as a result of hardware RAID, well, unless it's hardware RAID 0.

First, the HP LeftHand storage nodes can be configured in RAID 5, 6, or RAID 10, all with various useable capacities. This calculator only has RAID 5. Why choose? The reasons are many, but most common are performance, protection and capacity. You choose, it's no different with us or any other vendor solution.

Second, "Disk rightsizing", a term used to explain why that 1TB hard drive you bought only shows ~932GB useable. Why? Well, that's because the hard drive vendors view 1MB as 1000 kbytes while your Operating system views 1MB as 1024 kbytes. That extra 24 bytes adds up which is why you truly don't get the actual hard size useable (this is before formatting it with your favorite file system as well). Again, nothing specific to use or any other vendor solution.

Third, Network RAID. What is Network RAID? Network RAID is a unique feature of the HP LeftHand SAN that allows you to CHOOSE on a per volume basis how many replicated copies of your LUN / VOLUME are distributed across the SAN. What is unique about this is it's DYNAMIC. You get to choose which volumes have it and which do not.  What it offers you is the ability to survive entire node failures; if your nodes are physically separated and you lose an entire physical sites, your data remains online and available.

Hardware RAID is usually a set it and forget it configuration, and it's seldom changed. Network RAID is dynamic, because you as the customer choose to turn it on or off depending on the application protection needs. With that choice, you select the use of additional capacity to protect your data in a manner superior to standard hardware RAID. They key point here is choice. You have the choice, and if you change your mind, the system is dynamic and allows you to change the level of data protection on a per volume basis as often as desired. Unfortunately, a calculator without options isn't very reflective of real life. In essence, the above calculator was missing the infamous *Your mileage may vary...

To quote our previous blog poster "Unlike NetApp's space efficiency calculator, the LHN Duplication Calculator I've designed doesn't have any input fields or buttons..."

The HP Centralized Management interface WE designed, does have buttons, and even drop-downs, allowing you to choose how your capacity is used, ALWAYS.  

Tweet this! 


Posted 06-30-2009 9:37 PM by CalvinZ

Comments

Mathieu van Schaik wrote re: HP LeftHand capacity
on 07-01-2009 7:14 AM

If I understand your post correctly, NetApp did you a favor by taking "just" Raid-5 and Network RAID 2. Raid 6 or 10 and or Network RAID 3 or 4 would yield an even lower useable capacity?

(I acknowledge the benefits of replicated datasets and or better in-system RAID levels)

In the other post John Spiers wrote:

"Yes, Network RAID 2 can sustain any random 5 disk failure. In fact, a 4 array system configured in Network RAID level 2 can sustain up to 2 complete array failures, and up to 6 disk failures in each of the remaining 2 arrays, and all at the same time."

As this fits better in this post on capacity:

Can you explain how? What Raid-level is used within an array system? How many duplicates of data is required? What utilization will result in this config?

BTW: I fully agree with avoiding terms like "always", "never", "every time", etc.

Thanks.

Patrick wrote re: HP LeftHand capacity
on 07-01-2009 11:20 AM

One of my favorite catchphrase in customer meetings: "It depends on...". ;) Nice article.

Jasen Baker wrote re: HP LeftHand capacity
on 07-01-2009 8:31 PM

Thank you for your positive feedback! We want customers to understand the benefits various capacity utilizations offer. Everyone knows the negative benefits of utilizing more storage. What the customer cares about are the positive benefits if we make the choice to use it or not.

To answer the questions: "Can you explain how? What Raid-level is used within an array system? How many duplicates of data is required? What utilization will result in this config?"

There are many flavors, but let's break down the building blocks.

First, we designed the SAN to grow modularly using HP server hardware with built in redundant power, fans, disks, and nics.

Each "node" consists of either 8 or 12 drives, allowing the customer to choose the incremental growth size of their cluster. The 8 drive nodes can run in RAID 5, 6, or 10 while the 12 drive nodes can run in RAID 5+0, 6+0 or 10+0

Capacity used for each RAID level depends on the size of the drives, but here is the break down of utilization vs benefit.

Hardware RAID Protection

--------------------------------------

RAID 5 - 1 drive worth of capacity used to protect against any 1 drive failure.

RAID 6 - 2 drives worth of capacity used to protect against any 2 drive failures

RAID 10 - Half the drives worth of capacity used to protect against the first or second set of drives failing. (Aka, with 8 drives, 1-4 can be lost or 5-8 can be lost, but not one from both)

RAID 5+0 - 2 drives worth of capacity used to protect against 1 drive lost in drive set 1-6 and 7-12 but not more than 1 per set can be lost

RAID 6+0 - 4 drives worth of capacity used to protect against 2 drives lost in drive sets 1-6 and 7-12 but not more than 2 per set can be lost

RAID 10+0 - 6 drives worth of capacity used to protect against  all 6 drives lost in sets 1-6 or 7-12 but only one set can suffer failure.

Performance

-----------------

Raid performance consideration for each hardware RAID level can be very detailed, and articles on it have been written. So I would recommend looking at this Wiki link to understand the different performance details based on your hardware RAID selection here:

en.wikipedia.org/.../Redundant_array_of_independent_disks

Network RAID

-------------------

Network RAID is a simple selection of how many replicated copies of a volume do you want to have distributed across the nodes in your cluster. The higher the level of Network RAID you choose, the higher level of data protection that volume gets.

Network RAID level 2 allows you to lose every other storage node in your cluster and thus allows you to physically separate your nodes across buildings or even cities (assuming you have gigabit connectivity between sites).

The minimum Network RAID level is 0, which is a stripe of data across your nodes with no additional capacity.

The formula would be this:

Network RAID 0 = Original LUN Size

Network RAID 2 = Original LUN Size x 2

Network RAID 3 = Original LUN size x 3

Network RAID 4 = Original LUN size x 4

You can view a brief product demo which demonstrates network RAID at:

h71016.www7.hp.com/.../model.html|ProdPage|flash

Or better yet, download our SAN yourself using our Virtual SAN Appliance for VMware. This is a pre-packaged fully functional HP / LeftHand SAN that virtualizes internal storage running on VMware.

h20392.www2.hp.com/.../try.do

- Jasen / HP Storage Architect

Mathieu van Schaik wrote re: HP LeftHand capacity
on 07-02-2009 10:52 PM

Jason, thanks for your clarification!

Allow me a few remarks:

• “Network RAID level 2 allows you to lose every OTHER storage node”.  If I understand it correctly it is not allowed to lose 2 CONSEQUETIVE storage nodes. Node1 contains B1&B4, Node2 contains B1&B2, Node 3 contains B2&B3, Node 4 contains B3&B4. Suppose nodes 1 & 2 fail (unlikely, but it could happen)... no more B1  … downtime & data loss!

• Interpreting your explanation on Raid protection, in my opinion *GUARENTEED* without data loss & downtime within 1 storage node/raid-array are only:

o RAID  10:  1 drive. Suppose 1 disk of each set fails (unfortunately it’s not up to us people to decide which drives fail)

o Raid 5+0: 1 drive (it’s not up to us people ….)

o Raid 6+0: 2 drives (it’s not …)

o Raid 10+0: 1 drive! (… you know the drill)

• Unless I’m overlooking things, I’m just curious how this relates to John Spierings statement: "Yes, Network RAID 2 can sustain any random 5 disk failure. In fact, a 4 array system configured in Network RAID level 2 can sustain up to 2 complete array failures, and up to 6 disk failures in each of the remaining 2 arrays, and all at the same time."?

Alex McDonald wrote re: HP LeftHand capacity
on 07-06-2009 11:27 AM

Jasen

Thanks for the article.

You say;

"First, the HP LeftHand storage nodes can be configured in RAID 5, 6, or RAID 10, all with various useable capacities. This calculator only has RAID 5. Why choose? The reasons are many, but most common are performance, protection and capacity. You choose, it's no different with us or any other vendor solution."

I chose RAID-5 to give you the best start possible.

It is different with NetApp. RAID-DP (our RAID-6 equivalent) gives high performance, protection and capacity all at the same time. Our independent benchmarks like the SPC-1 are all done at higher capacity utlisation than other vendors, and well over the theoretical 50% capacity utilisation of RAID-10, or nRAID2/3/4. With thin provisioining and snapshots turned on too. www.storageperformance.org/.../benchmark_results_spc1

Charles wrote re: HP LeftHand capacity
on 11-02-2009 4:58 PM

I have 2 NSM2060s in a cluster.  Both nodes are set with RAID 5.  If I have a very large volume that spans multiple drives and one of the drives fail in that volume (with no replication level set), will the volume stay online?  If so will I have to simply replace the drive and let it rebuild?  If not, does the data become inconsistent on that volume and I will have to restore?  Thanks!  Great article.

Charles

Chris Brown wrote re: HP LeftHand capacity
on 11-03-2009 1:58 AM

Hi Charles,

The volume will stay online and you will just need to replace the drive and let it rebuild.

Thanks,

Chris

Add a Comment

(required)  
(optional)
(required)  
Remember Me?

Type the numbers and letters above:
Powered by Community Server (Non-Commercial Edition), by Telligent Systems