The first part of this comparison generated some heated interest in the blogosphere. It earned a response on Dell's Equallogic Blog as well as a couple of well composed comments from both HP and Dell. Dylan Locsin from Dell thought that a few of my points were inaccurate and he even insinuates that I may be misrepresenting myself. I'll respond to his points here:

1) Clustering: His point that my calling Equallogic "clustered" is inaccurate is completely correct. I should have call it "distributed." Equallogic does not offer true clustering across SANs. They don't offer a comparable feature to the Lefthand's Network RAID functionality which allows striping across SANs. Equallogic cannot pool IO resources as it scales.

2) Groups: Dylan interpreted my argument with Equallogic that "Data cannot span more than these two SANs" as a replication argument. It wasn't. With HP Lefthand SANs I can add performance and capacity with each node up to 16 nodes (and even beyond). This avoids silos of performance and storage. I did fail to point out that I was considering the PS4000 series which is the most commonly encountered Equallogic SAN considered device by my clients. The PS6000 series improves this situation.

Dylan wondered why I didn't offer a disclaimer that I sold HP Lefthand. If he scrolls down a bit he'll notice the disclaimer. I also prominently place the Lefthand SANs on my website. GCS does sell Dell servers, desktops and notebooks. We support MD3000i and Equallogic SANs from Dell. I think Equallogic is a fine product. However I firmly believe the HP Lefthand SAN offers the best value for my clients in the SMB market. That said, Dylan, I'm not opposed to also offering Equallogic to my clients that bleed blue. There are many of those, especially in Austin. I don't mean to insinuate that Equallogic is bad product. I've seen clients succeed with both options.

Joe has been posting a nice buyer’s guide about the features available on various SAN storage products. But it doesn’t really address the question of why move from direct attached storage (DAS) to a SAN. What benefits does a SAN bring to an organization?

To answer the question, we have to look at how storage is purchased and provisioned without SAN based storage.

Say you have an older File Server which currently has 500GB of data but can’t take any more drives.  Performance is fine but you decide you need to upgrade to a new server to add more capacity. You want the new server to have room to grow, so you spec out and order a server with 1TB of disk space. You also have Mail Server. It’s a pretty new server, but you recently merged with another company and the mail store size doubled overnight.  It needs storage and fast, but not a lot as you don’t anticipate the mail store growing so quickly in the future. You have 500GB of storage sitting on the new File Server, but you can’t use any of it to host the mail store.  So you order an external tray with just a couple drives and still end up with more storage than you need.  

Ten months later you look and see that file server data is growing rapidly and the free space is already gone.  You have an external tray attached the mail server, but you can’t use that to attach it to both servers. So you have to purchase a separate tray for the file server.

This example is over simplified, but it highlights three problems of direct attached storage:

·         Poor scalability

·         Poor Utilization

·         Silos of storage

Poor Scalability

It is not uncommon for an entire server to be replaced just to increase capacity. Not only is this expensive, but if the old server was performing fine you are buying additional resources (processing power and RAM) that you don’t really need. Adding external trays allows you to add large chunks of capacity, but what if you only need a bit more space. 

Poor Utilization

Data growth is dynamic which makes calculating your future needs difficult. The lack of flexibility in DAS discussed above and the fact that adding capacity frequently involves downtime or complete server migration leads organization to buy much more storage upfront than they need.  The result is servers with considerable amounts of unused storage.

Silos of storage

Because of the two problems above you end up with servers with spare capacity, but it is unavailable if another server needs it.  You have storage you’ve paid for but can’t use. Organizations respond in two ways, they either accept this inefficiency or start adding multiple services to servers that have space even if it violates best practices.  An example would be adding high usage file shares to your mail server.

How a SAN helps

A properly designed and deployed SAN addresses each of these problems.  SAN based storage is scalable, flexible, and allows storage to be shared between servers.  The end result can be better storage utilization meaning you only buy the storage you need and can easily add more as needed.  SAN based storage is also a critical element that enables organization to effectively utilize advanced technologies like clustering and virtualization.

Hopefully you now understand the benefits of SAN based storage and have decided it’s time to add it to your organization.  As you can tell by the number of features Joe has in his buying guide, there are many more options and decisions to make. Like any new technology deployment, good guidance can mean the difference between a successful deployment and failure. In future posts I’ll break down:

·         What is a SAN? Or more accurately what components make up a SAN?

·         How should a SAN be used and how should it not be used?  There are many ways to deploy a SAN can limit its utility and seem to be more expensive and troublesome than it works.

Most of our clients choose between an Entry Level or Mid-Level SAN, as described in Part 1 of this Buyer's Guide. In this post I'll identify the features that differentiate those two classes of SANs.

Clustering: Clustered SANs operate like clustered servers. If one fails the other takes over immediately. This is only available in the Mid-Level products such as HP Lefthand, Dell Equallogic, etc. Only HP's Lefthand offers true clustering out of the box. The minimum order for a Lefthand SAN is two completely separate units.

Thin Provisioning: Mid-level SANs allow  you to oversubscribe storage by allocating storage to a volume without reserving that storage. If you create a 100GB volume but only use 40GB, the remaining 60GB is free to be allocated to another volume. This feature is essential in maximizing the storage efficiency of a SAN.

Offsite Replication: Replication between SANs is the foundation for an excellent DR solution. Replicate all data and VMs to another site. Many of the mid-level SANs offer this solution but in some it is a licensed add-on. HP's Lefthand includes scheduled replication at no additional cost but real time replication and automated failover is an additional license fee.

Snapshots: Snapshotting technology is an on-array backup method that utilizes a relatively small amount of disk space. This is possible to restore entire volumes quickly without relying on external storage.

De-duplification: Long an enterprise only feature, integrated de-duplification is making its way into mid-level SANs. This can dramatically increase the efficiency of storage but can have a significant performance cost.

When should I recommend Cloud solutions to my clients? I don't know. Today I'm going to start figuring it out. I'm going to compare an on-premise build with a cloud based solution for a hypothetical 50 employee organization. Allow me to skip a lot of details and say it will provide roughly equivalent features and security. Anything required in both solutions was a wash and therefore ignored. What you're left with is mostly a lot hardware (servers, SANs, firewalls, switches for iSCSI, etc), deployment services and hardware-level support.

NOTE: All pricing is retail and rounded. Individual proposals may vary. A greenfield is assumed - no data migration. Everything that was common to both solutions was excluded so this may represent a small portion of the overall project. 

Cloud Offering:

In the cloud we'll procure 20Ghz of processor capacity, 50GB RAM, 3 TBs storage with 10 Mbps of bandwidth to host 15-18 Virtual Machines. Onsite backup for 1.5 TBs. This will run about $10,000 per month.

On-Premise Offering:

On-premise will provide provide about 60Ghz (over 24 cores), 96GB RAM, 3TB direct attached storage for backup, 3TB iSCSI SAN. We'll plan on hosting 25-30 VMs on this platform. You'll get this for about $160,000 including 200 hours for installation and configuration. See my build here. Note we dropped 25% of capacity for high-availability.

To even the playing field we have to factor in datacenter costs for the On-Premise Offering. I see a rack, 10Mbps of bandwidth, and 40A of power. Say $3000 per month?

Conclusion:

Let's compare 3 years of costs. In the cloud you're out $360,000. At home, including the data center costs, you're looking at $268,000. The $92,000 delta has to cover the hardware support and maintenance for 3 years. There is some difference in capacity and reliability. I hope it helps your evaluation to say that the cloud offering, in this case, is 35% more expensive.

Click image for single page PDF

When small and midsize businesses find the need to invest in a storage area network solution they are faced with lots of options.  The modular, clustered, iSCSI SAN products offered by EqualLogic (Dell) and LeftHand (HP) have become highly attractive solutions for organizations moving into virtualization that need robust solutions with integrated replication.  But, these are complicated solutions and gathering information to compare these offerings can be a challenge.

The Info Tech research group has posted a great PDF (registration required to view) comparing these two product lines and explaining what make clustered SANs different from the more traditional model.  For even more information, check out the blog equallogicversuslefthand.blogspot.com.