Archive | NetApp

Launch: NetApp Storage for Midsize Businesses

NetApp has just launched an update to the FAS2000 series for midsize businesses, along with some cool, new marketing collateral.

From the brochure:

  • Deploy virtual machines in minutes rather than hours and days
  • Cut backup time by 80%
  • Manage on average 500TB per admin
  • Use up to 83% less capacity for virtual machines
  • Achieve less than 5-minute recovery time objective
  • Achieve payback period in less than a year

I don’t normally talk about company product releases, but I found the launch video entitled “IT Survival Guide – Don’t Panic” to be particular amusing. Check it out!

FAS2000 Series Datasheet

Midsize Business Portfolio Brochure

Disclaimer: I am an employee of NetApp, but this is not a NetApp blog. Opinions expressed above are mine and mine alone.

5,490 total views, no views today


Quick Review: NetApp Flash Cache (PAM II)

I don’t normally do reviews on this site, but since there is a lack of real world information from customers regarding NetApp’s highly-touted Flash Cache (in the past referred to as PAM II), I thought I would make an exception this time.


Our environment consists of a FAS3160A (active-active configuration) with 4 disk shelves between the 2 filer heads. This is a fairly small environment with only 56 total disks, but were are running some intensive production workloads (i.e. Lotus Notes, MSSQL, MySQL, etc.) as well as some externally accessed Java application servers with their associated SLAs. We are also hosting small amounts of NFS and CIFS data in lieu of Windows file server VMs for some users. In order to conserve on disk capacity, we are also implementing de-duplication and thin provisioning on production workloads since we value a lean and efficient storage environment.

Of late, we have been having some performance issues. These consisted mostly of high disk latencies (averaging above 45ms) at the ESX hosts as well as some complaints from internal end users over occassional slow access and performance issues for a few of the services we provide.


Even though we just upgraded to the FAS3160A from a FAS3020 in February, I have been planning proactive performance upgrades since before that implementation aimed at keeping the performance at a high level. We ran into some scaling issues in the past so I didn’t want to repeat any of the same mistakes. I did a lot of analysis on storage performance and technical documentation prior to deciding that the Flash Cache should be a good option for our environment. Our read/write ratio averages about 10:1.

Coinciding too with the Flash Cache upgrade is a protocol migration away from iSCSI and to 8Gbps FC. I looked at going all the way forward to FCoE, but unfortunately our current blade chassis and interconnects only support an upgrade to 8Gbps FC in our current configuration. We will plan this for the next upgrade cycle.


NetApp and our reseller did a good job at making the price competive which is good, due to our limited budget for this hardware. I considered other options, but the Flash Cache gives us the best “bang for the buck”. I suspect not many customers are using this hardware in Japan, so perhaps we can serve as a case study for the benefits of Flash Cache when all is said and done. The ordering process was smooth and the hardware arrived with 15 days for installation.


The nice thing about having an active-active configuration is that we were able to add the hardware with absolutely no downtime! The first head unit tray(?) was removed after performing a cluster takeover and (2) 8Gbps FC cards and (1) Flash Cache 256GB card were installed and (1) Quad port NIC was removed. The same process was then repeated for the other filer head. I will use a similar procedure to update the ONTAP OS firmware at a date in the near future.


Prior to the Flash Cache installation, we were seeing consistent average disk latency of about 45ms across all 8 of our ESX hosts. This should be around 15ms or less to provide adequate response times for our VMs and finally to the end users. It was my goal to accomplish this with the Flash Cache as well as alleviate some of the high disk utilization that we were seeing (especially during backup windows) from all of the sequential disk reads.

The Flash Cache takes several hours to “warm-up” or load frequently used data into the cache. Since we performed the installation on Fri. night, I took my first look sometime on Sat. and the results were about what I expected.

Some of the rough quantitative results:

  • Reduced average disk latency to less than 25ms. This includes the local disk adapter which seems to affect the average. If I look at the specific disk adapter for the iSCSI devices, I see latencies of less than 10ms average.
  • Reduced disk utilization from 70-80% down to 20-30%. Less random reads on the disks frees them up for more important tasks (like sequential reads) for which read cache is less ideally suited.
  • Backup timeframes reduced by a factor of 4.


If you don’t want the expense due to increased energy/cooling/space requirements, the NetApp Flash Cache is a worthy alternative. Remember, the Flash Cache is laser-focused specifically on random read workloads (good news for most VMware environments), but do your homework and perform lots of tests and analyses to check (and double-check) the potential performance benefits to make sure it has a positive ROI for your environment.

Pairing the Flash Cache with an upgrade to a more modern, industry recommended protocol (1Gbps iSCSI > 8Gbps FC, in my example) can help keep the performance of the system ahead of the curve.

The Flash Cache also serves as good alternative, especially if you don’t require the additional disk space provided by adding additional spindles to the environment. This will likely be our next bottleneck now that the performance issues are under control… at least for the time being.

3,452 total views, no views today


Powered by WordPress. Designed by WooThemes