Gosh, is it October already? Last month marked my Seventh "Blogoversary". I started this blog seven years ago, in September 2006, to celebrate IBM's 50th anniversary of disk systems.
Several readers have expressed concern that I have not been blogging as much lately. For all my readers looking for a lame excuse, I just have two words: Jury Duty. Last month, I was selected for a specific trial. While many people dread the thought of jury duty, I found it a refreshing change of pace. However, I am glad to be back at work where I belong!
(For my readers outside the United States, jury service in USA is compulsory, Jurors listen to all of the testimony in a criminal or civil trail, ask questions, reviews evidence, and take notes. Thanks to a power called [jury nullification], members of the jury can disagree with the law the defendant has been charged with, and even reach a verdict contrary to the letter of the law, on the belief that the law should not be applied in that particular case.)
Continuing my belated coverage of the of the [ IBM Edge 2013] conference, I participated in the storage "Meet the Experts" panel, which is a long-time tradition, started at SHARE User Group conference, and carried forward to other IT conferences. The free-for-all is a Q&A Panel of experts to allow anyone to ask any question. These are sometimes called "Birds of a Feather" (BOF).
(Disclaimer: Do not shoot the messenger! We had a dozen or more experts on the panel, representing System Storage hardware, Tivoli Storage software, and Storage services. I took notes, trying to capture the essence of the questions, and the answers given by the various IBM experts. I have spelled out acronyms and provided links to relevant materials. The answers from individual IBMers may not reflect the official position of IBM management. Where appropriate, my own commentary will be in italics.)
|How should storage administrators deal with server virtualization?|
|We recommend you investigate the use of OpenStack. IBM storage systems like XIV, SVC and the rest of the Storwize Family support the OpenStack Cinder interfaces to provision block storage in support of server virtualization.|
|What are the interactions between SVC and Flash?|
Depending on which hardware model you have, SVC nodes can support up to four Solid-State Drives (SSD) each, a maximum of 32 drives in an 8-node cluster. IBM also announced the [ "IBM FlashSystem Solution"] that combines SAN Volume Controller (SVC) with All-Flash arrays, offering features like volume mirroring, thin provisioning, real-time compression and remote site replication.
(For more on this, see the IBM RedBook [Implementing the IBM SAN Volume Controller and FlashSystem 820])
|Can we get SONAS integrated with LTFS Enterprise Edition [ LTFS-EE]?|
|IBM has exciting developments in this area that it hopes to announce at a later date.|
Which should we use: XIV with SSD or go instead with the IBM FlashSystem?
|SSD inside the XIV provides extended read cache. If you would like to use IBM FlashSystem, you can put SVC in front of both XIV and FlashSystem, and use Easy Tier to manage between the two.|
|Can I put multiple XIV arrays behind an IBM FlashSystem?|
|FlashSystem is not an external disk virtualization device. Use SAN Volume Controller for this. (Certainly, SVC, XIV and FlashSystem is a winning combination!|
|Can we have multiple virtual switches on POWER7 nodes? Currently, my system only supports one virtual switch.|
|Perhaps you should ask that in the "IBM PureSystems" meet-the-experts discussion next door. You might need a higher level of VIOS.|
|Does IBM offer any tools to reclaim LUNs (orphaned volumes)? Our server and storage teams are separate, so this is a real problem for us.|
|Yes, IBM's SmartCloud Virtual Storage Center [ VSC], which includes IBM [ Tivoli Storage Productivity Center] software to identify LUNs with no I/O activity.|
|What is "Object Storage"?|
Unlike block-level storage, object storage is access through HTTP interfaces known as [RESTful APIs]. OpenStack offers [Swift APIs] for this. Many cosider such APIs as a pre-requisite for deploying Software-Defined Storage. Object storage may be less expensive by employing commodity hardware.
(For more on this, read [ Object Storage: The Future Building Block for Storage Systems -- A Position Paper] by Factor, Meth, et al.)
|Are five-minute intervals sufficient to determine storage performance problems?|
|IBM Tivoli Storage Productivity Center uses 5-minute intervals, gathering performance data from a broad variety of devices in your datacenter. Generally, this is sufficient for identifying and troubleshooting performance issues. If you need finer XIV-like granularity, you may need to use device-specific tools.|
|How can we bring processing closer to the data like with Oracle's Exadata?|
|IBM's [ PureData] and [ GPFS Storage Server] do this.|
|When will SVC, Storwize V7000 or DS8000 series products offer the Object Storage interfaces you discussed in the previous question?|
|IBM already supports OpenStack's Cinder interfaces for block-level access to storage, and is contributing as a Platinum Sponsor to the OpenStack for object-based Swift. Watch this space!|
|Rather than having to put separate ProtecTIER gateways in front of SVC or Storwize V7000, can we have SVC/V7000 just add the "Virtual Tape Library" protocol to its stack of host-attachment protocols?|
|Great idea! We will pass this on to IBM development.|
|With all of this "read acceleration" won't this increase the likelihood of a "write storm"?|
|Yes, write storms are coming, but can be controlled.|
|Are there plans to offer SVC behind SONAS gateways?|
|Yes, an iRPQ is available.|
|What is the biggest performance bottleneck for Flash?|
|Data moving through SAN switches adds only 5-8 microseconds of latency. Distributed systems often do not measure in sub-millisecond units, making it difficult to see improvements below 1 millisecond. Many performance issues arise from lazily-written applications. It helps to have Flash-optimized middleware, such as IBM DB2 BLU and WebSphere.|
|Since SVC adds 60 to 100 microseconds of added latency in front of IBM FlashSystem, is there a way to optimize the path through the SVC stack?|
|Existing parameters allows you to disable the SVC cache for particular volumes. We are investigating a more formal solution, a leaner code path for SVC with FlashSystem.|
|Are there any exciting enhancements to the [ SDDPCM] multi-pathing drivers you can share with us?|
|No, IBM is focused on MPIO multi-pathing instead.|
Thanks to all my readers who expressed concern over my lack of blogging. As you all know, [ blogging is like jogging], so getting back into the full swing of things requires extra effort on my part.
technorati tags: #IBMedge, blogoversary, jury duty, jury nullification, OpenStack, Cinder, Swift, RESTful, SAN Volume Controller, SVC, XIV, Storwize, SSD, Flash, FlashSystem, Redbook, SONAS, LTFS, LTFS-EE, Easy Tier, POWER7, PureSystem, PureData, SmartCloud, Virtual Storage Center, VSC, Tivoli Storage, Productivity Center, TPC, object storage, GPFS, GPFS Storage Server, GSS, ProtecTIER, SONAS, DB2 BLU, WebSphere, SDDPCM, MPIO