As readers of this blog know, I am a huge fan of RESTful management (1) because it is incredibly efficient when properly implemented, and (2) because REST itself is part of a suite of open standards, avoiding the lock-ins that too often accompany proprietary approaches to management. My favorite case study has been X-IO’s Intelligent Storage Element (ISE) technology, which is completely instrumented for management via REST.
Here are some screen shots of their current implementation…
ISE Systems Setup, manual entry of ISE list based on IP Address or DNS Name
This page contains the ISE List plus the new left-side navigation to any of the other ISE units in the list. This page also provides the navigation to the other new ISE Systems pages; Status, Storage, and Performance.
ISE Systems Status NOC View
The ISE Systems Status page provides both a NOC (Network Operations Center) View with easy to view pie charts for ISE units and their Components as well as a detailed Status Table View for each ISE in the list with status details with hyperlink navigation within the table. The left-side navigation tree also shows the “worst status to the top” rollup with hyperlinking available.
Towit, a more detailed ISE Systems Status Table View
The ISE Systems Status Table View provides individual component status per ISE with tabular column sorting capabilities, e.g., sort on Temperature, Power Supply, Controller, etc.
You can also look at how the storage is being used across a “field” of ISE bricks, with a Systems Storage page providing the high level info.
The NOC View provides pie charts to breakdown the total storage with Used versus Free Space and how the Used space is built using ISERaid1 and ISERaid5 across the “Field”.
There is also a pie chart indicating the breakdown of media types based on installed DataPacs (Hyper, Performance, Capacity, etc.). The horizontal bar chart gives the breakdown per ISE system in the “Field” with hover-over tool-tip information regarding Used and Free space. The “bars” are hyperlinks to take the user directly to “that ISE” Create Volume page to use more of the free space.
And of course, there are drill-down screens to provide detailed storage pool information for each ISE and for each Storage Pool with an ISE…
ISE Systems Storage Table View
And if you want to get performance info at a glance for one or more ISE bricks…
The ISE Systems Performance page provides the total performance across all the ISE as well as individual ISE performance metrics. The NOC view of this page gives the overall total IO/s and total MB/s across the “Field”. The horizontal bar chart shows the “per ISE” breakdown that makes up the overall totals. The ISE System Performance Table View shows the breakdown per ISE in a tabular form.
And the drill down…
Systems Performance Table View
Located below the horizontal bar chart, there are collapsed table headers, one for each ISE in the list. These table headers can be expanded to display the instantaneous performance metrics of the ISE and Volumes hosted on that ISE.
ISE Metrics View
The Individual ISE Metrics View contains the instantaneous performance data for the ISE which includes the Total IO/s, Total MB/s, a Table View with each Volume and metrics, the IO Load Balance between the two ISE Managed Reliability Controllers, the Read versus Write ratio, the Storage Pool IO Load Balance, and the individual volume Total IO/s with read and write breakdown in the vertical bar chart.
There is an up-to-2 hour IO/s performance line chart with Response Time for the ISE that is “browsed into”. The Total IO/s line in the chart provides a “hover” capability to display a larger, zoomed-in graph with a 10 sample range for more detailed analysis. The 2-hour performance history can be retrieved as a CSV exported file for use in other charting tools.
Okay, so enough dog-and-pony show. This rich information fount is readily available via any browser-enabled device using simple REST primitives. If you have other products that are also REST enabled — from switches to software — you could readily integrate it into an application centric or end to end infrastructure map that would facilitate coherent and universal management. Alas, the industry seems to have little interest in this idea at present. Everyone says they like REST, but many (like VMware) who have implemented it, place it behind proprietary APIs to ensure, I guess, that standards-based management does not reduce their wares to a component easily replaced by a competitor’s component that does the same thing.
Anyway, I like what X-IO has done with REST and I wish that every vendor would emulate their path. The cool thing is that the software stack for doing management via REST is published in the open at CORTEXDEVELOPER.COM and requires substantially less work to implement than does SMI-S or even proprietary APIs. The bad news is, REST lets you manage everything in common.
The next generation of X-IO management will see ISE rigs “friending” each other when they detect the deployment of another blade/rig/brick or whatever they are calling them this week. Imagine that bit of dialog: ”Oh, hi! You’re new here. I see that you are an ISE also. Want to share capacity? We can be BFFs.” Talk about storage simplified.
From Micro to Macro
Another thing I am tracking in the REST management world is CDMI from SNIA (but I won’t hold that against it). CDMI stands for Cloud Data Management Interface and SNIA says it is a standard that specifies a protocol for self-provisioning, administering and accessing cloud storage. To work, all participants (service providers) have to implement their storage in a compliant way. I’m guessing that a few providers have done so, but most haven’t. Also, I reckon that the granularity of what can be administered, configured, provisioned or monitored via REST also varies widely. Otherwise, we would no longer be talking about unmanaged infrastructure (the REAL problem that “private clouds” seek to address is poorly managed or unmanaged infrastructure, period.)
I am still researching CDMI to see about its current status and practical aspects of its implementation to date, probably for an article on ESJ.com. For now, I love this quote from a X-IO guy who demonstrated their REST implementation to SNIA when they were first trying to come up with CDMI, “a comment that came out of that original demo was ‘Wow, the export was built and the response came right back… if this was done with SMI-S, we would have to have gone for coffee before it would come back.’” Indeed, maybe SNIA should reconsider SMI-S for a REST-based approach.