Tag:

3PAR StoreServ

Squeezing the ultimate performance out of vSphere

by Philip Sellers June 2, 2014

written by Philip Sellers

Over the course of the past few months, I have had the opportunity to compare physical workloads directly against virtualized workloads. As workloads have moved into vSphere in the environment that I support, we have had the opportunity to compare workloads head to head, pinpoint some bottlenecks and remedy them in our environment. In addition, I have spent a considerable amount of time researching running business critical applications inside of vSphere.

Power Profiles Matter with Performance

In Elias Khnaser’s PluralSight course Best Practices for Running XenApp/XenDesktop on vSphere, he says to check the power management settings of your servers, particularly in blade servers, since they many come with a dynamic power profile based on processor performance states (P-states). P-states conserve energy in servers by operating the processor it different power levels. With a dynamic profile, it may be lowering power to the CPU in order to save energy, but at the expense of optimized performance.

hp_power_profile I checked this on my HP ProLiant blade servers and they were set to the default “Balanced Power and Performance.” The label doesn’t sound all that bad, but on further searching, this setting enabled the dynamic power management within a server. With this enabled, the CPU’s seemed to be powering down and taking additional time to power up when demand from vSphere increases.

Changing the Power Management setting to “Maximum Performance” ensures that the CPU stays powered fully, however, it may take additional power to do so. In my initial testing, this increased performance within my business critical applications where we are measuring transaction times. We were able to observe a decrease in the transaction times by nearly one half.

ilopowersettings

HP has a technical whitepaper outlining Power Regulation in ProLiant servers. I found two ways that you can adjust the power profile – both by entering BIOS during boot and also in the iLO on ProLiant systems. For ProLiant G7 and Gen8 systems, the iLO power profile change is instantaneous. For G6 and earlier systems with iLO 2, it appears to activate on next boot. I also found that depending on the profile setting on the hardware, vSphere presents different options at the hypervisor level. When the server is in dynamic mode, vSphere is able to configure a profile from the Configuration tab of the host. When set to “Maximum Performance”, the ability to change or set a profile is not available in vSphere.

In addition, VMware has a knowledgebase article (1018206) with troubleshooting steps for poor performance based on power settings. This KB article details not only HP ProLiant servers but also Dell servers.

And finally, there is one more place to check and that is the power profile settings within your Guest OS. Setting the profile to a high performance profile may also have a positive impact on performance.

Training: Business Critical Apps on vSphere

VMware has made some great online training available to all customers and the same courses are available to partners who seeking accredittation for virtualizing business critical apps (though Partner University). Each course is a 3 hour, online self-paced class.

Virtualizing Oracle Database on VMware

Paravirtualized SCSI driver

paravirtual2 VMware unanimously makes the recommendation to the the Paravirtualized SCSI adapters within guest OSes when running business critical apps on vSphere in all the courses linked above. Paravirtualized SCSI adapters are higher performance disk controllers that allow for better throughput and lower CPU utilization in guest OSes according to VMware KB article 1010398. The same KB article also gives step by step instruction for adding the controller and driver into Guest OSes.

Reserve the Entire Memory Allotment

reserveallguestmemory VMware recommends reserving the entire memory allotment for virtual machines running business critical apps. This ensure no contention where high performance applications are concerned. vSphere 5 and higher with virtual hardware version 8 or higher has a checkbox that allows for reserving the entire allotment of vRAM, even as allocations change.

Avoid Thin on Thin Storage Provisioning

From my own experience, running vSphere thin-provisioning works great on array technology that does not have native thin provisioning capabilities, but if the storage has thin provisioning at the array level, use that and instead use the thick provisioning in vSphere, however which thick provisioning option to use varies by array.

While moving an application onto vSphere in my environment, we had development setup on an array without thinning capabilities and production on an array with thin provisioning. The production environment took more than double the time to perform the data build-out when utilizing thin on thin capabilities. Once we noticed this, investigation showed that it was thin on thin that was causing the performance derogation. Once the disk allocated the blocks within the VMDK, the performance derogation disappeared. Our theory is that the slowdown occurs when allocating new blocks for data.

Converting the production virtual machine to thick disks by performing a storage vMotion to a new LUN caused the build to complete in less time than on the development VM backed by the array without thin capabilities. In my particular case, it was an EVA without thin capabilities backing development and a 3PAR with thin provisioned LUNs backing production. The recommendation for 3PAR is to use “Thick Eager Zero” disks since the 3PAR will recognize and zero out the space of the ‘eager’ zeroed disks. I have seen other recommendations for different arrays, where many recommend the “Thick Lazy Zero” disks.

Thin on thin also presents more challenges in estimating and accommodating growth in an array over time. Any thin provisioning technology should be monitored for over-allocation to ensure that there is adequate room for growth on the array. Doing thin on thin may obstruct this view somewhat, so proceed with caution and know that you’re likely losing some performance.

June 2, 2014 0 comments

Compute Datacenter Software Storage Virtualization

HP announces OneView integration with vCenter Server

by Philip Sellers February 10, 2014

written by Philip Sellers

oneviewheader

Today at VMware Partner Exchange (PEX), HP announced vCenter Server integration for its HP OneView systems management package. Using REST API calls to HP OneView, the vCenter plugin is now able to initiate actions in HP OneView directly from the vSphere Web Client. HP OneView, introduced last September, is a central platform of the Converged Systems group and it is the glue that takes disparate components and makes them into a cohesive solution. The current version has support for HP BladeSystem, Virtual Connect and Gen8 ProLiant blades and rack-mount systems. Proliant G7 blades are also supported. It has a roadmap for adding support with HP networking and HP storage in the future. HP OneView provides both deployment, management and monitoring capabilities so that administrators can build out solutions and maintain solutions.

HP Grow Cluster - vCenter with OneView

As part of a demo of the integration, HP is showing off the ability to completely automate the build and deployment of ESXi nodes in a new cluster and into existing clusters. By utilizing OneView’s engine, a VMware administrator can kick off the orchestration of a new hardware build directly from vSphere Web Client using a simple “HP Grow Cluster” menu item (pictured above). The menu item kicks off a wizard drive workflow (at right) that calls the tasks in OneView via its REST API. The vCenter plugin allows the administrator to pick a reference server profile, create multiple server profiles from the template and then uses HP Server Provisioning to deploy an ESXi host build plan.

At the core of OneView is the software-defined model of IT – where the underlying hardware is abstracted from the software-based configuration and settings. The reference server profiles define a base level of hardware and combined with settings, including firmware levels, BIOS settings and the ProLiant server model and generation preferences. Once the reference profile is deployed, it becomes a server profile assigned to physical hardware. Like Virtual Connect server profiles, these are portable and can be migrated between physical hardware.

Customers who are familiar with HP’s Insight Control for vCenter will notice many similarities between the OneView integration and the Insight Control integration. The OneView plugin for vCenter shows network diagrams of port mappings and configuration as well as end-to-end storage mapping from the VM to the LUN and storage array. The current iteration of OneView relies on the same Insight Control for vCenter storage plugin. OneView has yet to adopt storage control and management directly into its platform, though it seems safe to expect its coming given HP’s direction.

The majority of this announcement is that VMware administrators can now kick off OneView tasks directly from the vSphere Web Client, monitor and view the progress and completion of the tasks it spawns. OneView is still in early in its life, but the interface is simple and intuitive. Its not overly complex at this time and its build with modern concepts, like search, deeply integrated into the core product. Extensions like the vCenter plugin are great enhancements for administrators who are accustomed to working in a native platform and gain access to the capabilities of OneView directly from a familiar console. This release continues on HP’s track record of delivering usable functionality directly to VMware administrators from vCenter.

The HP team has a video demonstration of the HP Grow Cluster feature in the vCenter Web Client posted at hp.com. It is definitely worth a watch to see the entire process orchestrated with just a few clicks and settings.

February 10, 2014 0 comments

Datacenter Storage

Exploring HP’s Power of Common Architectures in Storage

by Philip Sellers January 31, 2014

written by Philip Sellers

During HP Discover last month, I had a great chance to hear from David Scott, Senior Vice President and General Manager of HP Storage, and also to have a conversation with Siamak Nazari, the Chief Software Architect of 3PAR. One of HP’s talking points for 3PAR for the last several years has been around the idea of Polymorphic Simplicity – a marketing term that they use to describe how HP is positioning the 3PAR architecture from the mid-range, to performance to high-end storage all with a single architecture, feature set and interoperability.

With the current arrangement of 3PAR systems, HP has the mid-range StoreServ 7200 with 2 nodes max, the StoreServ 7400 with 4 nodes max, it has the all-flash performance focused StoreServ 7450 and then the high-end StoreServ 10000 series arrays with a max of 8 nodes. All of the systems share the same 3PAR ASIC and software. The management interfaces and configuration concepts are the same regardless of which model. The underlying OS is the same across all – meaning that although the 7000 series was introduced just 1 year ago, its runs the same OS as the flagship 10000 series arrays which have been proven. The illustration below is from HP to show how the 3PAR architecture fits many different use cases where different factors are most important – all with one common architecture underneath.

HP-One-Architecture

While I’ve heard the message at previous HP events, it wasn’t until talking with Siamak Nazari that I began to understand why this has been such a big point from HP. Evolving a storage OS is a difficult tasks for any company. That’s why storage innovation often happens with start-up companies that build a new design from the ground up, instead of inside of mainline vendors evolving their dual-controller arrays into a modern storage array. HP’s purchase of 3PAR brought them a modern storage architecture that was natively multi-node capable. Where the complexities are exponentially more difficult to move from single controller, to dual controller to multi-controller, HP took the opposite approach by adapting a modern 8-node architecture down to a 2-node and 4-node configuration.

The real key for HP 3PAR customers is that by HP building a portfolio that spans a wide breadth of use cases, it expands the install base of the overall architecture while refining and enhancing the underlying OS. Enhancements made for one particular use case get shared by all of the customers through shared enhancements. Customers ultimately win by getting incremental improvements from their existing investment – all delivered by software.

HP has been consistently shipping improvements through software on the 3PAR StoreServ systems for several years. When it introduced the HP 3PAR StoreServ 7450 in June of 2013, the all-flash array boasted OS improvements for handling all-flash drives in the array. In December, HP revised the 7450 and nearly doubled performance to 900,000 IOPS with less than .7 milliseconds of latency [100 percent Random Reads, 4K block size local to controller] and more than doubled the possible capacity by introducing a new technology called Adaptive Sparing and by introducing new 480GB and 920GB SSDs at lower price points.

3par-ssd-dec2013

Adaptive Sparing allows the HP’s 3PAR StoreServ array take over coordination of sparing flash sectors instead of having the individual drive handle sparing. HP found that it could manage the sparing capacity as a pool more efficiently than by letting the individual drives handle this — and that allowed the 3PAR StoreServ arrays to access more of the actual flash capacity. The net result is an increase in overall utilization and lower the price point for flash in the array.

So, where is the benefit here for all? While the Adaptive Sparing technology was developed specifically for the all-flash 3PAR StoreServ 7450, the same Adaptive Sparing technology is also benefiting any HP 3PAR customer who is using SSD. This was one of the best examples of benefits from a common architecture that I had heard from HP.

But common architectures is not just a benefit from the HP 3PAR StoreServ – it is an approach adopted by other HP Storage solutions, also. HP’s StoreOnce D2D backup, StoreVirtual arrays and StoreAll NAS hardware are all based on their ProLiant hardware configurations. Since HP has made the conscious decision to build out solutions based on ProLiant servers, each generation provides performance increase and innovation that is also shared with software solutions. During Discover in Barcelona, Spain, last month, HP announced the new generation of the StoreOnce – the StoreOnce B6500 – based on the ProLiant Gen8 server hardware which boasts considerable performance gains obtained both by software and hardware improvements.

Taking the common architecture further to the software stack, StoreOnce and StoreAll also have a lot of shared DNA within their software stacks. Both solutions leverage the IBRIX technology that HP acquired in 2009. According to David Scott, the IBRIX distributed file engine is used across a couple product lines, although not specifically talked about. While HP knew that the IBRIX technology was fantastic technology, it was really about targeting the correct solutions for a problem and tuning the software to specific solutions. And like the shared ProLiant and shared 3PAR technology, incremental improvements on one side will mutually benefit the other sides as well.

January 31, 2014 1 comment

Compute Datacenter Software Storage

HP Discover Barcelona: What to expect & how to follow online this week

by Philip Sellers December 8, 2013

written by Philip Sellers

I am currently sitting in Myrtle Beach airport waiting for my first of three flights that will ultimate bring me to Barcelona, Spain. I’ll be spending the next week attending HP Discover in Barcelona. It is one of HP’s two major conferences each year and it will be first visit to HP Discover in Europe and on a personal note, my first trip to Europe in thirteen years.

So what to expect from HP Discover?

Some of the usual highlights are expected – the keynote from CEO Meg Whitman to address customers directly and general sessions for HP executives to unpack their vision and product lines for customers and highlight some of the successes the company is achieving. There will be hundreds of breakout sessions for onsite attendees. The Discovery Zone is the place to visit with hundreds of HP’s top technical staff available to answer questions and explain solutions in detail. Unlike other conferences, HP staffs not just marketing folks in the booth, but also engineers and developers. But if you’re not onsite, keep reading – you can still follow the conference.

What else is expected?

Storage: Last year’s HP Discover brought major news with the introduction of the HP 3PAR StoreServ 7000 series arrays. These ‘baby’ 3PAR arrays have been a big hit in the mid-market storage marketplace. Their introduction converted me and my employer into HP 3PAR users. I’ll be interested to hear what’s new in 3PAR. In June at Discover, HP announced the 3PAR StoreServ 7450 array – an all flash variant of the already successful 3PAR StoreServ 7400. But HP 3PAR is just part of the HP Storage story – and there will likely be other things going on around software-defined storage during the week.

Systems: HP created the HP Converged Systems group in May. This August, during VMworld, HP held a reception for bloggers and media where we were briefed by the newly formed group and the Converged Systems group is responsible for full system solutions made out of HP intellectual property, similar to VirtualSystem, CloudSystem and AppSystem solutions we’ve seen delivered to date. This will be their first chance to show fruit during the week.

Servers: HP Servers group was also announced in May as HP combined its Business Critical and Industry Standard Server groups. It will also be the first show for this group since their combination this year, since June’s HP Discover happened immediately after the announcement. In previous years, there have been a lot of bleedover between Business Critical and Industry Standard, back and forth. From the industry standard server side, HP ProLiant Gen8 continues to roll out models to complete its refresh cycle and some ProLiants are now in their second revision. HP Discover in June brought the announcement of the HP ProLiant MicroServer Gen8 which I have since had a chance to review in depth. Excited to see what’s new here…

Mobile: I’m also interested to see what more HP is doing in the mobile space. In the past year, HP has been toying with Android based hardware, Windows 8 based tablets and has other groups working on enterprise mobility solutions. One solution that I saw in June – called HP Anywhere – was in a early beta form. It allowed developers to have a secure platform and SDK for developing rich HTML5 applications for Enterprise. Mobile is a growing space and I am curious what HP has been developing in this space to help enterprise retain control of data, secure and enable users.

But I’m not there – how can I participate?

I know most people who read Techazine will not be on-site at HP Discover Barcelona. But, that doesn’t mean you can’t participate. Here are some opportunities for the next week for your to participate online:

HP Storage Press Event – Monday, 9am EST – http://hpstorage.me/1buyzV4 – Register online and attend for free to watch the announcements from HP Storage live.

HP Discover Opening Keynote featuring Meg Whitman, CEO, and Bill Veghte, Executive Vice President and General Manager, Enterprise Group – Tuesday, 8am EST – http://bit.ly/1bQrhNz
HP Discover General Session featuring George Kadifa, Executive Vice President, HP Software; Mike Nefkens, Executive Vice President, Enterprise Services; and Dion Weisler, Executive Vice President, Printing and Personal Systems – Wednesday, 8am EST – http://bit.ly/1aLQfr9
Live and on-demand videos all throughout the week on a variety of topics – http://bit.ly/1fa5lf3

On Twitter, follow hashtag #HPDiscover
Follow @HPStorageGuy – Calvin Zito’s HP Discover Barca Bloggers list on Twitter – https://twitter.com/HPStorageGuy/lists/discover-barca-bloggers and talk with us – ask questions and interact.

Hope to hear from you!

December 8, 2013 0 comments

Datacenter Storage Virtualization

VMware releases VSAN and the power of policies

by Philip Sellers September 19, 2013

written by Philip Sellers

VMware Virtual SAN (VSAN)
Diagram by VMware

VMware is pushing back on storage vendors as it seeks to show the world the full power of software-defined storage. A couple weeks ago, VMware announced the VMware Virtual SAN (VSAN), a new software-defined storage platform tightly coupled with vCenter server. The new platform requires both hard disks and solid state disks to create a pool of redundant, shared storage within a cluster. The hard disks are the core of the storage and the SSD are used for buffering and read caching but do not contribute to the total amount of storage. And although the name has SAN in it, the traditional SAN that is bread and butter of its parent company, EMC, could not be more different than VSAN.

In many ways, this is similar to the VMware Virtual Storage Appliance, but VSAN is much more integrated into vCenter and the vSphere platform. The requirement of SSD is also new and will greatly contribute to overall responsiveness for the storage. But more importantly, the new software defined storage platform is an example of the possibilities yet to come from tightly integrating storage with vSphere.

The real power of what VMware has released is not the storage technology itself or the simplicity that it can be deployed (single-click simplicity like HA and DRS). The real power is the addition of Storage Policies completely throughout vSphere and VSAN. The policies are the real capability that can be leveraged across multiple types of storage. As I said earlier, I think VMware is simply showing all storage vendors this is what is possible. The VSAN is certainly not a SAN replacement for most enterprise customers, but there are certainly remote office or small office use cases where this will fit nicely without the need of purchasing a purpose built storage array — use cases, by the way, where virtual storage appliances (VSA) currently fit.

But back to the policies. Introduced in vSphere 5.0, Storage Policies allow administrators to tag and set classing of storage for different pools that are available within a vSphere cluster. The Storage Policies allow an administrator to abstract the back-end storage and provide a list of capabilities to end-users or operators who are creating new virtual machines – and choose a policy during initial deployment to meet their needs. You see, what VMware has discovered is that the end users don’t care if the back-end storage is a 3PAR or a VMAX – a Synology or a Drobo. What they care about is a class of service provided, response times and how those translate to their application and its performance. They also care about the cost of charge-back, if an organization has implemented those types of billing scenarios. Policies are all a way of defining those attributes back to the end user and abstracting the actual storage from those decisions.

I am sure many VMware customers are like myself. When I started the virtualization project, most systems ran on local storage. The SAN was reserved for our most mission critical systems and most of those were clustered. Today, however, vSphere has moved all of those workloads onto the SAN. It has increased my resiliency and ability to survive a disaster, but its also driven storage costs up exponentially. Yes, VMware is owned by EMC – goal to sell disks accomplished.

But the VSAN must run a bit counter to the EMC culture, so back to my original point. In many ways, I think the VSAN is simply a proof of concept of what is possible with software defined storage. It is a way for VMware to show the rest of the world what we can do. VMware is very good at fostering its ecosystem and partner vendors, so I think this will spur new innovation.

And so, for a third time, back to policies. With the introduction of VSAN, VMware is extending this capability of Storage Policies to allow administrators to assign a policy to a VM on creation and then the policy lives with the VM allowing policy based management of the VM over time. Not only can it be setup initially with the VM, it can follow and adapt as the back-end storage changes over time. With the addition of storage pools and maintenance mode for storage, VMware has automated a lot of the manual labor around managing storage. By creating the pools of storage and moving datastores in and out of them, administrators and move or remove storage without interrupting service of virtual machines. And also by applying tags and policy around those storage pools, it allows for the addition and removal of datastores as required to onramp or offramp storage as purchasing or lease cycles dictate. All of this while maintaining service levels and agreements for customers and workloads – whether those customers are internal or external to your organization.

Beyond backend maintenance of storage, say the attributes of a particular application or VM changes over time. Say it now requires a higher tier of storage – well thats an easy policy change as well on the VM and with the magic of storage vMotion and fully-automated storage DRS, the vSphere platform will perform the necessary moves to increase performance and move the workload onto the appropriate storage de jour.

In many ways, this is all done with the idea of reducing the menial tasks that an administrator has to perform. Its mean to kill the tracking spreadsheets and other mechanisms we have traditionally relied upon for classing storage or moving workloads appropriately. It all goes to making the storage layer of vSphere smarter and more automated.

Now, the last piece of policy that really sets VSAN apart from the competition. Within VSAN, the vSphere administrator can dictate the type of RAID, the number of stripes and copies that are present for a particular VM or datastore. For the first time, the core vSphere platform is controlling storage down the stack and automating how it operates. vSphere policy now defines where and how replication occurs, how and where stripes of data are implemented, and on a per VM basis, quality of service attributes to prevent noisy neighbor problems. And combining those together gives context and visibility up the stack in vSphere, too. The two way communication of the policy and the storage implementation means that the administrator can more granularly dictate how the storage should act to support the requirements of the virtual workload.

So, I have to say, after a week in San Francisco at VMworld 2013, the VSAN is the one piece I am most excited to test and evaluate. If for no other reason, you should sign up for the VSAN beta to see where the future of storage is going. I think the writing is on the wall of every vendor is talking software-defined, the VSAN represents the first software-defined storage product with such a deep level of integration into vSphere. I fully expect to see other vendors like Nexenta and HP StoreVirtual quickly adopting this paradigm – maybe even making it to market with product offerings around the same time that VSAN becomes generally available. API calls and other integration points should be very simple for software storage implementations. I’m excited to see where this next step takes us.

September 19, 2013 0 comments

Datacenter Storage Virtualization

VMware unwraps vSphere 5.5 – What’s in it for your Datacenter?

by Philip Sellers September 9, 2013

written by Philip Sellers

Almost two weeks ago now, VMware unwrapped vSphere 5.5 to the public. The release, which ships later this year, includes a lot of enhancements and some new features for the datacenter. I won’t lie – 5.5 has a lot more in it for the datacenter than 5.1 enhancements. And I left VMworld 2013 much more excited about the direction of the virtual datacenter than I did at the end of VMworld 2012. So what are the things that sys admins will really care about in vSphere 5.5?

Expected enhancements – bigger maximums and more Web Client

As expected, the new release of vSphere increases maximums. The ESXi host maximums were doubled from vSphere 5.1 to vSphere 5.5 in all cases, except for the maximum LUN size, which got a surprising 32X bump in limit to 64TB. The number of logical CPU’s doubled to 320, the number of virtual CPU’s doubled to 4096 and the amount of RAM has doubled to 4TB per host. Those represent some pretty incredible numbers. vSphere 5.5 also introduces Virtual Hardware version 10. The new hardware version is introduced to support newer CPU architectures as well as a new virtual-SATA Advanced Host Controller Interface (AHCI) with up to 4 controllers and 30 devices per controller. So, the new AHCI supports 120 total devices connected to a single VM.

Also expected, vSphere 5.5 drives users more towards into the vSphere Web Client for management. All new capabilities are controlled from the vSphere Web Client. This follows suit with vSphere 5.1, whose new capabilities like shared-nothing vMotion, were available only in the Web Client. One thing I am hearing repeatedly on social media is that the vSphere 5.5 Web Client is noticeably more responsive, which is great. I am also hearing the new Virtual Hardware 10 VMs can only be managed through the Web Client, which could cause heartburn for stand-alone hosts, but the final release may change this functionality. The beta version of the Windows VI Client throws an error if you try to edit a Virtual Hardware 10 VM. The error says the version 10 VM must be managed through the Web Client. This just means that resiliency of the vCenter servers continue to be critical parts of virtual deployments as it is a required part of managing the infrastructure.

And speaking of vCenter Server, the vCenter Server Virtual Appliance (VCSA) continues to become a more capable alternative to a Windows server deployment of vCenter. With vSphere 5.5, the VCSA has a retooled embedded Postgres database capable of supporting 100 hosts and 3000 VMs, although those numbers may actually change for better or worse. During one Ask the Expert session during the week, a member of the VMware Labs team said they were using a single VCSA to support as many as 500 hosts, but mileage my vary. This is one of those numbers that may change during final release.

I’ll file SSO enhancements under expected as well. While most people outside of VMware did not know what VMware intended to do with SSO, one session at VMworld pretty much summed it up by saying that the new version of vSphere SSO releasing with vSphere 5.5 will be what should have been released the first time. To many adminstrators, myself included, the version of SSO introduced with vSphere 5.1 just didn’t seem up to VMware standards. It seemed to be a third party tool that was included but not fully streamlined into the vSphere implementation.

The vSphere 5.5 SSO version includes a number of streamlined improvements. AD integration is much improved and includes the ability to tie it to multiple forests, which is needed for many larger environments – or anywhere a merger has occurred. There is no longer a need for an external database. And there is a multi-master replication model built into this new version.

Storage improvements

There are many storage improvements included in vSphere 5.5. One that I am very happy about is the streamlined and scriptable UNMAP process used to identify and clear out free space on a VMFS datastore so that your storage array can re-thin and free the space on the back-end. This is a big one for me with 3PAR. I want to keep my LUNs on a diet and this new ESXCLI command will go a long way towards helping me do that. In addition to the new command line, the back-end functionality has changed with UNMAP, hopefully making it more acceptable to run during production hours and moving towards a time when automatic UNMAP during deletes might return to default functionality. UNMAP can now stagger the reclamation process so that it doesn’t saturate a backend array with IO.

Another nice feature, from an administrative standpoint, is the automatic removal of permanent device loss (PDL) state feature. This is especially handy in the event of array failures, which have been known to happen. In the event of a failure, automatically clearing a LUN in PDL will keep the administrator from having to perform rolling reboots to completely clean up the mess. The handling of PDL conditions has improved greatly over the versions I have used from 3.0 forward. There as a time that a PDL could completely hose an ESX host and a reboot was the only way to recover.

vSphere 5.5 also features enhancements around SSD and Flash storage. The ESXi hosts now have the ability to work with add-add and removal of SSD devices, allowing a graceful handling of SSD storage as it becomes increasingly more common. In the world of Flash, a new read cache feature is introduced allowing for the aggregation of Flash storage into a pool that can be consumed by VMs for the purpose of accelerating read intensive workloads. Last, but not least, 16Gb Fiber Channel support, end to end, is available in vSphere 5.5.

Network enhancements

In the networking world, vSphere Distributed Switches just got a needed boost of capabilities. Traffic filtering has been added allowing administrators to filter the types of traffic. QOS has been added to the mis allowing for prioritization of traffic from the vSwitch up to the physical switch layer. Security officers and admins troubleshooting will both be happy to see the addition of a host level packet capturing tool that is the equivalent of tcpdump to Linux admins. 40Gb NIC support is also being introduced with vSphere 5.5.

Mission critical workloads

One of the message points from VMware during VMworld 2013 was about making vSphere a platform that can run any mission critical workload. The maximum enhancements and other resiliency features are all targeted towards that goal, but in particular one additional capability was introduced with vSphere 5.5 that is meant to assist with virtualizing latency sensitive workloads, like VOIP. A new Latency Sensitivity setting is now available on VMs which help reduce latency for necessary virtual workloads. When set to a ‘high’ setting, the hypervisor actively tries to reduce latency by reserving memory, dedicating CPU cores and disabling network feature prone to latency.

Tying it all together…

There is really so much more in vSphere 5.5 – a couple great places to get more detail is available from Kyle Gleed in his A Summary of What’s New in vSphere 5.5 post and in the What’s New in vSphere 5.5 document on the VMware website.

All in all, I left the Monday morning keynote of VMworld 2013 feeling a little let down. Many of the enhancements that will really improve the admin’s life didn’t get any stage time or mention during the keynote – one reason for the delay in posting details – I had to find details… Two of the big deals that were given stage time this year were the VMware VSAN, a software defined virtual storage implementation baked heavily into vSphere, and NSX, the network virtualization platform that VMware acquired with Nicira. I hope to post on these two in depth in the coming weeks.

I was very happy to see a bit of a return by VMware to its core audience and products – centered around datacenter virtualization. I had the opportunity to talk to a lot of other bloggers and industry folks while I was at the conference and I found several who shared my view that VMware had strayed from its roots in years past chasing the cloudy dream — a dream that we felt was a bit beyond the masses. And while cloud adoption is growing, most enterprise datacenters are still reliant upon VMware and its vSphere platform to running daily operations. Its good to see that VMware did not abandon us all and that they are are still working very hard to build the best and most resilient datacenter platform available.

Questionable decisions in years past – like the nasty 4 letter word – VRAM – can now be jokes and chuckles. With so much attention given to vCloud and programming layers in the past couple years, it was welcoming to see things added to vSphere, where I live and work. In my opinion, this year VMware had to deliver something valuable to all of its customers to starve off competition and justify the continued spend for Service and Subscription (SnS) contracts. In many ways, I think they delivered. I’ll save my thoughts on where they still miss for another post.

September 9, 2013 0 comments

Datacenter Storage

vSphere 5.5 introduces new, simpler UNMAP command

by Philip Sellers August 26, 2013

written by Philip Sellers

Welcome news for anyone using array-based thin provisioning with vSphere, VMware has introduced a new and simpler UNMAP command that can be invoke from esxcli with vSphere Platform 5.5. This will help 3PAR customers more easily perform the UNMAP process to free space on LUNs so that the array can discover and free the space on the back-end. This is a big deal for any thin-provisioned LUNs in a vSphere environment. In addition to an easier way to invoke the UNMAP, VMware has also made changes in how the UNMAP occurs on the array.

According to VMware’s What’s New document, the previous method of creating a temporary file and then informs the array the the blocks in the temporary file can be reclaimed. In the case of the 3PAR, the zero detect process then runs to identify the free space and reclaim it. With version 5.5, the reclaim allows the administrator to specify the size in blocks rather than a percentage and now the dad space can be reclaimed in increments instead of all at once – to help shops that experience performance degradation with the old method.

The new command is run from esxcli, logically under storage and VMFS.

[code]esxcli storage vmfs unmap[/code]

The full syntax of the command is not clear yet and I’ll have to wait for my 5.5 upgrade to test or get more details.

The fact that the new UNMAP is run from esxcli makes me hopeful that I will be able to script and schedule the UNMAP to occur during non-peak hours when the maintenance would not be service affecting.

What is UNMAP and what does it do? See this post for more information.

August 26, 2013 1 comment

Datacenter Storage Virtualization

Keeping VMFS datastores thin on 3PAR StoreServ

by Philip Sellers July 31, 2013

written by Philip Sellers

Earlier this year, I moved a large part of my infrastructure onto HP’s 3PAR StoreServ 7400. As promised, our initial migration netted us over 50% savings through the thin technologies baked into the 3PAR StoreServ platform. But one my biggest concerns was how well could 3PAR maintain its lean status over months of operation with my vSphere running there. (Looking for how to free space on 3PAR, jump down to 2 Commands to Reclaim section below.)

After several months of running on 3PAR, I can report that it does a noble job identifying and thinning the storage under my vSphere environment. But, its not perfect. 3PAR can only identify what’s truly zeroed and on a VMFS5 datastore, that equates to space that has not been utilized. vSphere doesn’t clean up its VMFS datastore in vSphere 5.1. And this behavior has nothing to do with 3PAR.

History of Space Reclamation in vSphere

With vSphere 5, VMware introduced a new primitive using the SCSI UNMAP command to its VMware APIs for Array Integration (VAAI). Billed as VMware’s Block Space Reclamation feature, it allowed vSphere to communicate with the storage array and identify which space was no longer occupied by a VM and could be reclaimed. VMware shipped vSphere 5 with the EnableBlockDelete advanced storage feature enabled by default.

Unfortunately, the Block Space Reclamation feature induced overhead on the arrays and caused varying performance impacts on systems, in some cases causing severe performance degradation. Ultimately, VMware decided to disable the EnableBlockDelete advanced feature by default and disabled is the recommendation for production environments. Fast forward to vSphere 5.1 and VMware has not reversed its recommendation.

So where does that leave customers with bloated filesystems? The good news is that you can run a manual UNMAP using the vmkfstools command on a particular LUN. Effectively, the manual UNMAP writes a large balloon file on the filesystem which is used to zero out the space that was previously occupied so that the storage array can identify it as free space.

On 3PAR, once the vmkfstools command is run, 3PAR then begins its discovery of freespace through the Zero Detect feature and over the next several hours will begin reporting true free space.

2 Commands to Reclaim

The command is easy for initiating an UNMAP in vSphere. But, you have to do a bit of prep work. UNMAP is supported with 3PAR StoreServ arrays running HP 3PAR OS 3.1.1 or later. See VMware KB article 2014849 for the details on how to confirm to determine support for other arrays.

You will need to enable SSH on the ESXi 5.0 or 5.1 host.
To do so, go to the Host and Clusters view in vSphere Client, choose the host from the tree on the left side, choose the Configuration tab in the right side, choose Security Profile under the Software section, then click Properties… under the Services section. Scroll down to find SSH, double-click and then start the service. ESXi will display a warning that SSH is enabled. When you complete running your commands, don’t forget to disable SSH.
SSH into the host. Navigate to the VMFS datastore you would like to perform the UNMAP on.
Once you attach to the host using SSH, issue

[code] cd /vmfs/volumes/<datastore_name>[/code]

Now that you’re locate inside the LUN, you are ready to perform the UNMAP. The actual command is simple.

[code]vmkfstools -y 99[/code]

Using the -y parameter, you specify the integer percentage of free space to reclaim. If the LUN has 300GB of free space reported in vSphere, the command will create a balloon file that is 99% of 300GB in size (297GB). Once the file is created, it is removed zeroing the space. You should adjust your percentage as you see fit. Starting somewhere in the 80’s would be a good place to start.

The safest recommendation is to run the storage reclamation outside of business hours or at least outside of peak business hours. Impacts may vary, but in my experience the 3PAR handles the requests easily .

July 31, 2013 2 comments

Datacenter Storage

HP enhances 3PAR StoreServ Peer Persistence with automatic failover

by Philip Sellers July 3, 2013

written by Philip Sellers

VMware Metro Storage Clusters running on top of 3PAR StoreServ storage have gotten an added boost of automation. HP announced 3PAR OS 3.1.2 MU2 during HP Discover two weeks back and one of the features that it introduced is a quorum witness node to assist in Peer Persistence installations. Peer Persistence is the HP 3PAR StoreServ’s version of metro clustered storage for VMware, providing transparent failover of storage from one array to another. With the introduction of the quorum witness, the administrator no longer has to start the failover manually and the witness server adds monitoring and remediation in the event of an outage.

The quorum witness is a separate virtual or physical machine that monitors the availability of both arrays in each site. In most configurations, it should sit in a third site monitoring the two 3PAR arrays. If one of the sites goes down or offline, it detects the status and begins failover on the remaining 3PAR StoreServ array. Because the 3PAR StoreServ uses ALUA to block and open the active IO paths, the failover is seamless to the VMware workloads. All of this is accomplished by using the same virtual volume world wide name on both arrays and by blocking IO to the secondary copy.

Source: HP

During HP Discover, I was able to talk with several HP engineers about the new feature and one of my initial concerns was whether this is an all or nothing configuration for automated failover. Fellow blogger, Justin Vashisht, confirmed that each virtual volume or group of virtual volumes can be controlled automatically or manually – the administrator has the choice when configuring the volumes.

When we purchased our 3PAR StoreServ arrays earlier this year, our sales team confirmed that the automated failover would be forthcoming and so, as a customer, this is a feature we are happy to see delivered quickly. One of the unknowns to us was how HP would architect the quorum witness, but in talking with Siamak Nazari, an HP Fellow and chief software architect of HP Storage division, HP adapted the 3PAR Storage Processors to also do the duty of quorum witness. This means the new feature is based upon what should be solid and field-tested code.

Nazari and David Scott, HP Storage Senior Vice President and General Manager, stopped by the bloggers area at HP Discover for a coffeetalk with a large group of us. One of the things that I found most interesting from the talk is that HP has somewhat future-proofed the 3PAR StoreServ arrays it is already shipping – by baking in additional capabilities into the ASIC and by activating those with software at later time. That leaves hope that there will be other new capabilities introduced for existing customers, although nothing specific was revealed to us.

When I first discussed the Peer Persistence feature on my blog earlier this year, I praised the simplicity of the solution both from an architectural and administrative standpoint. While advancing the features in the new release, I think that HP has also made a wise choice to keep things simple even with the quorum witness added to the mix. As an administrator, I appreciate the straight forward and easy to understand (and troubleshoot) approach of the solution.

Another welcome feature for GUI fans, HP has introduced control of Peer Persistence under the Remote Copy tools within the 3PAR Management Console GUI. This means that administrators now have the choice of configuring and initiating failover from within the GUI, though I think I still prefer the single command-line command.

And as of the time of writing, HP reports that 3PAR Peer Persistence is VMware Metro Storage Clustering certified, although I do not see it in the VMware compatibility matrix yet. Lots of great incremental improvements to the 3PAR StoreServ line with this new OS revision.

Disclosure: HP sponsored my trip to HP Discover, covering all travel expenses, however HP does not control the content of the posts that I write based on my attendance to the show. The thoughts, impressions and information contained in this post are my own.

July 3, 2013 1 comment

Datacenter Storage

HP Storage is touting unified architectures for the future

by Philip Sellers June 12, 2013

written by Philip Sellers

Its all about unified approaches for HP Storage as they push forward into the future. David Scott, senior vice president and general manager of HP Storage, has spent a lot of time during HP Discover talking about HP’s unified approach in storage architectures as opposed to competitor’s fragmented approaches.

In the StoreServe and StoreOnce product lines, HP has been working to create a full portfolio of product offerings based around a central technology – in the StoreServ’s case – the 3PAR ASIC, in the StoreOnce, based around the HP Labs StoreOnce deduplication technology, and with the StoreVirtual with a pure software approach. It has been working to deliver solutions of different shapes and sizes around these common architectures. HP calls it Polymorphic Simplicity, or multiple solutions meeting different needs around a central architecture.

3parpolymorphic

One of the key benefits of the common architectures approach is the ability to manage products with a common interface and management style, meaning less training and learning curve for administrators. In the 3PAR StoreServ lines, administrators get the same GUI and command line interface, the same OS and feature sets across all levels of the product – spanning from entry-level to enterprise. With its StoreOnce backup and recovery portfolio, HP has packed the same deduplication technology into its backup product and into its portfolio of disk to disk backup devices. In addition, all of the devices run a common OS and have a common management interface. The other key in polymorphic simplicity is data portability between solutions – achieved through Peer Motion on the HP 3PAR arrays and through deduplicated replication on the StoreOnce.

HP’s New Product Announcements

In December, HP introduced the HP 3PAR StoreServ 7200 and 7400 at HP Discover in Frankfurt. The 7200 and 7400 were meant to address customer situations “where value matters” and the existing 3PAR StoreServ 10000 class was meant for tier 1, mission-critical needs. But, as Scott says, this left a gap in between “where performance matters.” So, HP has introduced the HP 3PAR StoreServ 7450 – an all flash array to meet customer needs when performance is key.

The 7450 is capable of 554,000 IOPS (4K random reads) with .7 ms latency. This is achieved through a specialized ASIC which takes into account flash characteristics. The ASIC tunes IO and cache flushing specifically for flash. Thin provisioning and system-wide striping native to 3PAR help address flash wear issues to extend the life of the flash storage by saving writes that are unnecessary and by spreading necessary writes across all of the disks in the array.

In addition to the new hardware offering, HP also announced Priority Optimization being added to the 3PAR OS across the entire portfolio. This feature will allow for guaranteed control of IOPS across a multi-tenant array. IOPS can be allocated at the tenant and application level allowing for SLA-type guarantees or simply partitioning and isolation to occur within a single array. The 3PAR line also received integration enhancements for Hyper-V and OpenStack Fiber Channel support.

HP also unwrapped the software-only StoreOnce VSA, marrying software defined storage with information protection. The StoreOnce VSA allows customers to run the StoreOnce on virtual platform not exclusively dedicated to backup and recovery. Like its dedicated hardware siblings, the StoreOnce VSA supports the StoreOnce Catalyst protocol and deduplicated replication across all of the product line. It is architected to be multi-tenant capable, allowing for backup and a service from service providers using the StoreOnce. Some of the use cases include remote office and branch office deployments in the enterprise utilizing existing hardware and small-to-mid business who don’t want dedicated storage hardware but want to leverage standard servers with the StoreVirtual VSA and StoreOnce VSA for primary store and information protection.

June 12, 2013 2 comments