Storage

New storage project, part three: Budget difficulties and researching future solutions


In two previous storage articles this year, I talked about my new organization's storage dilemma -- that we need more storage plus a new backup system on a tight budget. I also provided a list of priorities that should be met by a new solution.  To recap, here is what I was looking for, in order of priority:

  • Block-level shared storage that easily connects to servers
  • 100 percent redundancy and able to withstand the loss of any single component
  • Reasonably priced
  • Snapshot capability
  • Replication capability for disaster recovery

Since I had to balance needs with cost, I was willing to sacrifice some items to meet my budget; for example, if I found a solution for a price that just couldn't be beat, but it meant no replication, that would be okay.

Although "reasonably priced" was third on my list, it quickly became the number one factor in my decision. I regret to say that this budget year does not allow for my new storage project. I'm forced to push this new project to the next fiscal year and stick with the status quo -- direct attached storage -- for now.

Never fear, though! I did at least get to the point where I had found some interesting potential solutions. So until I am able to begin this storage project in earnest, I'm going to share my thoughts on three possible solutions and discuss their pros and cons. Do bear in mind that these are only potential solutions and would definitely not be suitable for every environment. However, I believe that all three have a niche, depending on the needs of an organization.

What would you do with this storage challenge?

I also want to take this opportunity to ask you for your input. Let's make this a group project. I welcome you to respectfully comment on my selections, with the understanding that these are products that bear further scrutiny. Tell me why you think a particular solution is better than another, or why you believe a particular solution is less than ideal. Finally, I invite you to share possible solutions that are not included here. Please provide some justification for your choice. I will assemble the best comments and suggestions for the last installment of this series. For that, I will investigate the compelling solutions that you provide and compile as much information as I possibly can for your perusal.

First, here are the three solutions that caught my attention:

  • EqualLogic Peer Storage Array
  • Left Hand Networks
  • Dell AX150 (not the Dell AX150i)

My short list includes mostly iSCSI-based units, although the Dell AX150 is a Fibre Channel unit. For many organizations, iSCSI continues to provide the most value for the dollar. Although iSCSI is seen by some as technically inferior to Fibre Channel, I believe that with the right hardware, iSCSI can be a real contender, even when it comes to raw performance.

All three of the solutions above meet my initial three criteria: block-level shared storage, redundancy, and price. Even though all units are reasonably priced, there is still a wide gap between the Dell (lowest) and the EqualLogic unit (highest). The pros and cons below will shine some light on this. Note that I am not allowed to provide actual pricing that was extended to me for these units. Further, none of the vendors were aware that I was writing this series.

EqualLogic

The last time I went through the selection process, EqualLogic's iSCSI solution (a PS200E) was the clear winner. In my analysis this time around, their product has only gotten better. The PS300E provides 7.5 TB of raw storage while the PS400E provides 10.5 TB. Further, the solution is fully redundant and will work with nothing more than just a couple of good gigabit Ethernet switches. I will be the first to admit that I have a bias toward EqualLogic, but with good reason. The last time I selected an EqualLogic solution, I was more than impressed and the unit operated exactly as promised by the company. You can't get much better than that!

  • Pros
    • Excellent performance based on previous experience
    • Best overall redundancy solution: Each individual unit can be 100% redundant.
    • Excellent snapshot capability (no additional cost)
    • Excellent replication capability (no additional cost)
    • Easily expandable
  • Cons
    • Highest overall price per TB
    • Not as linearly expandable as the LeftHand solution: It's easy to add space, but not as easy to do so in smaller chunks. EqualLogic does sell units with less capacity (as little at 1.75 TB raw), but these are not inexpensive.

LeftHand Networks

Lefthand Networks is another contender in the iSCSI market. LeftHand makes its own hardware and arrays, but it also OEMs its software to other hardware manufacturers, such as HP and IBM. For this loose analysis, I was looking at the LeftHand NSM 160, which is a unit that houses up to 2 TB of raw capacity. Three units, or a total of 6 TB of raw space, create a fully redundant solution.

  • Pros
    • Excellent performance based on actual testing: This is, in part, due to the smaller, more plentiful units. Each additional gigabit Ethernet connection adds additional transport capacity to the storage solution.
    • Fully redundant with the right hardware selection
    • Excellent snapshot capability (no additional cost)
    • Excellent replication capability (no additional cost)
    • Easily expandable: I like LeftHand's expansion possibilities more than EqualLogic's. With a LeftHand solution, I can add space in smaller chunks, thanks to the way that LeftHand sells hardware. This provides a more incremental upgrade path than is reasonable with EqualLogic. EqualLogic does sell units with less capacity than the PS300 and PS400, but they are still fairly pricey.
    • Good price
  • Cons
    • Redundancy: I know it sounds contradictory. Part of the reason that LeftHand's solution is more incrementally expandable than EqualLogic's is due to the fact that the solution's redundancy requires multiple storage units, each with a fraction of the capacity of the total pool. In contrast, EqualLogic sells a unit that is fully redundant, but with the total array capacity limited to a single device. (You can add additional devices, of course, but we're talking about a small storage project here). However, it's hard to let go of that "hardened" mentality, where you want every possible device to be capable of suffering a fault. With LeftHand, it's necessary to change that mindset to every possible service that is able to suffer a fault and continue operating.

Dell AX150

Dell's entry level Fibre Channel SAN -- which is a rebranded EMC unit -- is a good choice for small organizations that have fairly limited storage needs.

  • Pros
    • Outstanding price: lowest cost per TB. (Read: Cheap as heck!)
    • Redundant with optional second controller
  • Cons
    • WYSIWYG: This unit tops out at 6 TB raw and is not expandable. The EqualLogic and Left Hand units form a storage pool that is expanded as additional units are added.
    • Limited configuration possibilities: RAID 5 and RAID 10 only. I would like to see a RAID 50 option as a middle ground. RAID 5 may not provide quite enough protection, but RAID 10 requires significant disk space overhead.
    • Tops out at 10 total logical server connections. However, when used with VMware ESX, a connected ESX host server can be configured to count as only a single server connection, even if there are multiple VMs on the ESX server. This fact is not true of the AX150i, which is a solution that I would not consider for my needs.
    • Requires a Fibre Channel switch to go beyond four physical server connections.

It's obvious that these possibilities are not all in the same class. The Dell AX150 is easily considered an entry-level solution while the LeftHand and EqualLogic solutions can be easily considered enterprise-grade solutions. The AX150 would only be useful for small organizations or for specific applications for which disk space requirements can stay within the limits of the device. With the best protection -- RAID 10 -- the AX150 would have only 3 TB of usable storage.

For EqualLogic and LeftHand, some storage folks will look down on iSCSI. However, it's rare that the data transport mechanism -- iSCSI or Fibre Channel -- is the bottleneck in all but the most data-intensive environments. Further, as I've written before, iSCSI is a great choice for most.

Summary

Once budget money becomes available, I will get to the final steps in the process and walk through the installation. In the meantime, I would love to hear about some of your success stories as well as your horror stories. The final part (for now) in this series will be based on your experiences and solutions.

About

Since 1994, Scott Lowe has been providing technology solutions to a variety of organizations. After spending 10 years in multiple CIO roles, Scott is now an independent consultant, blogger, author, owner of The 1610 Group, and a Senior IT Executive w...

5 comments
zipdogs
zipdogs

Scott, Nice work on the analysis. I recently read and article by Howard Marks, "The Truth About Storage Reliability." He discusses the studies by CMU and Google on the replacement history of more than 100,000 disk drives in data centers. The bottom line is that disk drives are far less reliable than what's published by the drive companies, and as drives get bigger rebuild times increase significantly, which opens the door for a second drive failing during a RAID re-build. I believe this is why you see RAID-DP from NetApp and EMC just announced RAID 6 for Clariion. When looking for redundancy, controller MTBFs are so much higher than disk drive MTBFs that it almost becomes a non-factor, plus most controllers fail because the cache battery dies, which can easily be replaced, and there is no data loss because they have a safety mechanism that places the controller in write-thru mode when the battery gets low. The AX150i doesn't offer double disk fault protection in a RAID set and neither does the EqualLogic box. LeftHand's clustering does protect a customer from double disk fault protection. I also understand that their architecture allows you to lose complete disk shelves and sites without SAN volumes going offline. There is a capacity penalty, but if you split a cluster across sites you are actually getting sync rep for free without having to purchase software, processor heads or additional disk shelf capacity. If you look at it this way they actually have the best capacity utilization story for muti-site SANs. Good luck with your project. John

gordonmcke
gordonmcke

You should check out Compellent's StorageCenter product too.... It will be more expensive than these three SANs mentioned, but has lots more expandability and functionality. As always, I enjoy your posts, Scott.

mfarley
mfarley

There is some reason to suspect that this comment was made by John Spiers, CTO of Lefthand Networks. -Marc Farley, EqualLogic employee.

allen.d.west
allen.d.west

Scott, I work for a very fiscally conservative manufacturing company in St. Louis, and I have done a similar study for my company over the last year or so. At first management did not want to purchase a SAN due to the high initial investment. However, I was able to prove to them that over the next 4-5 years we will save at least $30,000, and reduce our number of servers & racks if we implemented VMWare VI3 along with the SAN. In addition, we would have all the extra features of the SAN and VI3. I did a lot of research and compared iSCSI SANs from Equallogic, Left Hand, NetApp, and EMC. We ended up choosing Equallogic and are going to order our new SAN this week. I must say that some of your previous posts helped influence our decision. In my research, I could not find any negative information about Equallogic, and their only cons are the ones you listed in your article. I am going to try to get as much disk space as we can afford (either PS300 or PS400) so that we will not have to upgrade any time soon. My Equallogic rep even gave me a demo unit to try out for two weeks. No one else has even offered to do that. It took less than a half hour to get up and running and it performed very well. I pounded thoroughly it by running IO stress testing tools like IOMeter concurrently from two servers with three dual-port NICs each. Any of these SANS would have worked well in our environment, but we decided that Equallogic iSCSI SAN was the best solution for us because we are a small IT staff, and we always have a lot on our plates. They are so easy to set up and administer and Equallogic?s support is supposed to be the top notch. From my experience, their pre-sales support has been the best I have ever had. I personally would not recommend the Dell branded EMC hardware. Been there, done that. If you are thinking about buying any of the low-end Clarion-based EMC products, then read this article: http://searchstorage.techtarget.com/magazineFeature/0,296894,sid5_gci1258336,00.html If you are still interested in EMC, then go up to a CX3-10 which has both iSCSI and FC connectivity. We received a quote for just the CX3-10 SAN and no additional software (ie snapshots, replications, etc), and it was about 25% less than what we were quoted for an Equallogic PS300. I just think that the low-end Clarion EMC SANs are based on older technology. The CX3 series has pretty much the exact same features of the FX series EMC SANs I worked with five years ago, except that they have an iSCSI controller attached to them. Sorry to hear that you have to wait until next year to get your SAN. The silver lining is that prices keep falling and iSCSI technology keeps getting better, so you will get more for your money next year. Keep up the good work with your posts & articles. Allen