During the Joint Cloud Lab Olocity and Xiotech demonstrated the use of CDMI with Representational State Transfer (REST) web services for accessing and controlling storage resources through the cloud with Olocity’s StorageIM and Xiotech’s Intelligent Storage Element (ISE) and CorteX (RESTful Web Services) products.  The demonstration illustrated how products using industry standards and open source can easily interoperate together to provide cloud services. The below diagram is an overview of the demonstration.

Upcoming Demonstration Events:

Don’t miss seeing Olocity and Xiotech demonstrating real product implementations of  CDMI and REST for providing cloud storage resources and services at the upcoming events below:

• OGF 29 http://www.ogf.org/OGF29 and HPDC http://hpdc2010.eecs.northwestern.edu conferences in Chicago, IL from June 20 to 24, 2010
• FLGISA 2010 Annual Conference http://www.flgisa.org/Conferences.aspx in Clearwater, FL from July 20 to 23, 2010
• SNIA Storage Developers Conference http://www.snia.org/events/storage-developer2010 in Santa Clara, CA from September 20 to 23, 2010

For the books, I’m a advocate in storage standards. I’m also an advocate for vendors thinking outside the box and that is exactly what Cortex is about. Cortex, Xiotech’s ‘Embedded’ RESTful storage API for their Intelligent Storage Elements.

For starters Xiotech chose a RESTful web service that has implemented XML as its current mime type and utilizes HTTP methods for monitoring, provisioning and indications. The URI’s implemented for capturing information about the ISE are similar to the SMI-S view concept.  As an example, the cortex ‘query’ URI call (localhost/query) looks familiar to the SMI-S Multiple Computer System Profile. The difference is you don’t have to create any association traversals.

Cortex also has a set of XSD’s that we used to dynamically create our client object model. These definitions made it a snap to create a normalized DB schema that performs and scales nicely for trending and utilization.

All of the above being said I also never touched the ISE element manager during development. I provisioned everything through simple http URI posts and would check to ensure endpoints were created, volumes were created through simple http URI gets.

During and after implementing our initial ISE aggregated monitoring solution a question kept creeping up, why are storage standards so complex?  Xiotech has completely removed any complexity.

The challenge from our perspective is determining from a heterogeneous point of view the following:

• What vendors support standards and to what extent
  • SNIA SMI-S
  • Web Services
• If SMI-S is supported, what version of SMI-S does that vendor and hardware models support?
• If SMI-S is supported, what vendors support statistics and in what SMI-S profiles do they reside?
  • What vendors support the SMI-S Block Server Performance (BSP)?
  • What ‘specific’ vendor’s hardware supports the BSP capabilities?
  • What ‘specific’ BSP metrics are available for each vendor’s hardware model (e.g. Top Level Computer System, Component Computer Systems, Ports, Disks and Volumes)?
• How does one determine the above and test ones development against each and every hardware platform?

What has become clearly apparent after years of working with standards is that unfortunately most end-users do not recognize the SNIA SMI-S standards and how it should assist them in determining their hardware and aid in avoiding vendor lock-in. From our perspective during the development of these core metrics it also became apparent what vendors are invested in SMI-S and what vendors really have not sunk their teeth into standards or have rejected standards completely.

Below list what we have chosen as our 1.0 performance metric based on our test customer requirements:

Storage Array

• Support for component computer system port statistics
  • IOPS
  • Kbytes Transferred / sec.
• Support for volume statistics
  • IOPS
  • Kbytes Transferred / sec.
• Support for disk statistics
  • IOPS
  • Kbytes Transferred / sec.

Switch

• FC Port Statistics
  • Bytes Transmitted
  • Bytes Received
  • Packets Transmitted
  • Packets Received
  • Link Failures

Results

The below screenshots are sample StorageIM performance charts of a Symmetrix array and QLogic Switch utilizing the SNIA SMI-S. In my opinion EMC has done a superb job constantly extending their SMI-S implementations and ensuring that as their product lines grow their implementation of standards grow alongside. The primary switch manufacturers have also made sure their product lines implement standards as their product lines grow.

We’ve also had the opportunity to implement these new features at sites around the globe. We were actually quite amazed at some of the feedback.  Below is the feedback we received from one of our test customers. We’ve taken out the name of the storage vendor and their product to protect the innocent.

“I really like the improvements on the hardware stats.  The numerical values for the IOPs and KB/sec is exactly what we’ve been looking for.  I’ve already been able to identify our busiest LUN, which we were not able to do before given the tools we had.”

What is becoming quite obvious is the lack of visibility of performance metrics for storage administrators.

I will also note that quote above also mentioned lack of SMI-S performance statistics for the storage vendors Disk Drives. This is a definite issue with the user citing a need and use case, “Last week, we had so many disks fail that it didn’t rebuild one of the arrays, and while we were noticing degraded performance,  we had no indication of why until the storage vendor’s technician showed up with disks in hand…”

If you are interested in testing StorageIM we would be more than happy to give you a 60 day evaluation.

Screenshots

StorageIM Storage Processor Port Statistics

StorageIM Disk and Volume Statistics

StorageIM Switch Statistics