My slide deck from the recent webinar series 24 Hours of PASS, Summit Preview session, where I presented a session called Virtual CPUs: Right to Ludicrous Speed, are now available for you to view up at SlideShare. Let me know if you have any questions! I hope to see you all at the PASS Summit later this year!
I get asked all the time – “why virtualize SQL Server if I’m content with my physical servers today?” The normal answer, outside of the usual answers of increased agility, flexibility, cloud-readiness, etc., is that license optimization is possible if your organization has enough scale to license a set of virtualization hosts and manage VM density to maximize licensing.
But, what if your organization does not have the sheer volume needed to cross that break-even point on the virtualization investment? What if the SQL Server version, platform, and app are happy? What other talking points do we have that can encourage virtualizing every SQL Server?
There are quite a few less-tangible benefits, such as VM migrations for hardware or SAN upgrades, ease of system-level backups and disaster recovery, improved operational high availability choices, and greater flexibility and agility in the datacenter. However, it’s difficult to put a dollar amount on this one. All of these are operational benefits that save staff time and energy. But, what about saving cost?
We need to look a bit further out. Eventually, the third-party vended app vendors are going to stop supporting the older versions of SQL Server, and the new versions have all migrated to core-based licensing, which can potentially lead to an unexpected cost during application upgrades. Moving to VMs now – or then – allow you to allocate the number of vCPUs that are needed at the time, rather than whatever the current hardware contains. Future physical server purchases will get harder to find smaller core count servers, all of which will increase the license spend. The vCPUs in a SQL Server instance can be licensed by virtual cores, and that can contribute to a reduction in future licensing purchases. The fact that these VMs could reside on the current VM infrastructure (as long as it is capable and contains enough free resources) could also help reduce or eliminate a future hardware purchase as well.
I know I’m preaching to the virtualization-friendly choir here, but if SQL Server virtualization is not the right call for right now, as time goes on, the necessity for maintaining or reducing CAPEX along with reducing OPEX will make virtualizing these systems a more appealing option as time goes on.
And… virtualizing these platforms is step number one in your cloud readiness strategy. And you are moving towards the cloud, right?
If you are actively managing VMware environments with workloads that have high performance needs (such as all of the virtualized SQL Servers that we work on), this new book called VMware vSphere 6.5 Host Resources Deep Dive, written by Frank Denneman and Niels Hagoort, is a must read book! Designed for VMware-oriented system architects, this book walks the reader through each layer of host resource allocation and management in a way that no other book has ever accomplished.
This book redefined the phrase “deep dive”, and I’m taking a lot of notes for future authoring and presentations. The depth of content is unparalleled in tech authoring.
Topic sections such as CPU, memory, storage, and networking, the four main areas of resource management on any virtualization host, are presented. Anyone can talk about these settings in general, but the advanced VCDX-level topics are covered in incredible depth. We work with a lot of advanced virtualized SQL Server VMs, and significant discussions from this book for performance tuning these SQL Servers include:
- Advanced vNUMA balancing and optimization
- CPU core counts versus clock speed
- vNUMA memory speeds and non-local memory access
- Clearing up misconceptions about vSphere Balanced Power Management
- Queues and resource allocations
Go get this book – NOW! It’s a must read, and read it twice. No, three times. Then give the book to colleagues who can benefit. I’ve got two copies on my desk right now, and know who these will be sent to after I’m done re-reading them!
I’m proud to announce that I’ve been selected to present a session in this year’s 24 Hours of PASS: Summit Preview webinar series entitled Virtual CPUs: Right to Ludicrous Speed. This session is scheduled for July 20th, 2017, at 03:00 GMT, or July 19th at 11pm Eastern.
These webinar series are a sneak peek at some of the sessions to be presented at this year’s PASS Summit in Seattle the last week of October. These sessions are from world class speakers and authors who we have all been studying under for years, and now you can see them live!
Session abstract: One of the largest points of contention with virtual SQL Servers and the VM administrators is how to configure the CPUs. Experience says more CPUs are better for performance. VM admins say less is better. Third-party vendors say you need all of them (and it doesn’t matter how many your hosts have either). Can over-provisioning virtual machine CPUs speed things up, or does it slow things down? What is the right methodology to determine the correct number of virtual CPUs? How does this configuration align with the physical servers? From sampling and analyzing performance data, to “right-sizing’ your SQL Server virtual machine CPU count, to properly aligning the VM with the physical server NUMA topology, you will gain the understanding of how to properly manage and validate your virtual SQL Server vCPU configuration in this insightful session. Valuable tips and tricks will be shared that you can take back to your virtual SQL Servers and immediately apply to your own environments.
I’m proud to announce that my sessions for this year’s VMworld 2017 conference in Las Vegas have turned up in the VMworld Content Catalog! I have four exciting sessions (if you’re a database geek) at the conference, and for those DBAs with VMware administrators attending this conference, tell them to attend these sessions so they can help build a better platform for your data.
The second revision of our business critical applications and databases pre-conference workshop is to be held at VMworld 2017 US this year on August 26th!
We will cover all pertinent aspects of best practices for deployments and ongoing management of MS SQL Server running in virtualized infrastructure. This deep-dive full-day workshopwill be delivered by VMware technical SQL Server specialists working in concert with world renowned external SQL Server and Virtualization experts, including me! The attendee will experience a workshop rich with technical content to include subjects such as vMotion for Failover Cluster Instances, Availability groups, SQLaaS with vRealize Automation and Site Recovery Manager. The content will be focused on best practices for design, implementation and management augmented with anecdotes of successful customer implementations.
SQL Server professionals, listen up. If your infrastructure admins are attending VMworld this year, please let them know about this course. Training them in the intricacies of the VMware platform as it relates to SQL Server, and having the training come from SQL Server professionals, is going to make your virtualization experience better.
Register for VMworld here! Add the training course at the bottom of the VMware Education courses during the registration process.
with Michael Corey
VMware vSphere 6.5 has made a lot of changes to enhance the performance of Monster virtual machines (VMs). Databases by their very nature are the classic Monster VMs. If Monster VMs are not virtualized properly, they will never perform well and can negatively impact the performance of other VMs on the hosts. In this presentation, we will teach you how to properly virtualize Monster VMs/databases with vSphere 6.5. We will discuss why you should virtualize and take a look at some installation issues as well as how to architect for performance in terms of vSphere 6.5 specifics, including the storage layer, the processor, memory considerations, and the network layer. Topics will also include NUMA, memory reservations, and how to avoid common mistakes. These lessons will help you optimize any workload you are virtualizing.
with Thomas LaRock
Business-critical database platforms are the last holdouts to enterprise virtualization. These systems are the most resource-demanding and latency-sensitive applications, and at the first symptoms of infrastructure challenges the DBAs blame the infrastructure. Successful virtualization of these platforms requires a different approach to virtualization from other applications. The speakers will share their years of experience in virtualizing data platforms. You will learn how to validate your VMware environment and manage the performance properly. Scalability concerns will be addressed through discussions on scaling database VMs upwards and outwards as you work with the DBAs to manage the data lifecycle. Key talking points will help you learn how to perfect database virtualization from a technical, an organizational, and people level.
Continuing from last year’s successful session, learn from some of world’s most renowned SQL Server experts’ real-life experiences gained while virtualizing the most demanding SQL Servers! In this panel discussion, you will also have the chance to ask your own questions about SQL Server on VMware vSphere and get answers from the SQL Server gurus.
For those of you who don’t manage SQL Servers in a VMware environment, check out all of the fantastic sessions in the performance tuning areas, as they are sure to help you boost performacne across your vSphere farms.
Register for VMworld USA today, and make sure to select the SQL Server on VMware boot camp as part of registration! I look forward to seeing you all there!
VM snapshots are one of the best virtualization features ever. But…have you ever had a VMware vSphere or Hyper-V snapshot grow out of control and fill the datastore the VM resides on? You know what happens… even the best VM admins out there seem to get burned once when they create a snapshot for some routine maintenance and just forget about it afterwards. All VMs on that LUN crash and go into a suspended state until free space is added to the datastore.
Sometimes this task is harder than it sounds. If your SAN is out of space, or the SAN management tools are out of your control, you could be stuck.
But… follow a simple trick to give yourself that last little bit of wiggle room in the event that a snapshot fills a datastore.
Add a large text file to the root of the datastore that you can delete if you need headroom! I know it sounds too simple… but it’s simple and effective.
These directions are updated for VMware 6.5 environments, and work all the way back through vSphere 5.5, but the same concept applies to any hypervisor out there today.
If you’re on a SAN that performs some sort of data savings, through either compression and/or deduplication, this trick also consumes next to no space on the SAN.
How do you do this?
Enable the SSH server, and remote into one of the ESXi hosts. This example is from a vSphere 5.5 ESXi server from my home lab.
Change directories into the root of the datastore that you want to create the file in.
Use the Linux command ‘dd’ to write a 10GB file (or whatever size of your choosing). Basically, we’re creating a file full of zeroes with a 1MB block size and ten thousand blocks. Simple, eh?
dd if=/dev/zero of=snapshotfailsafe.txt count=10000 obs=1M ibs=1M
It will take a few minutes to create the file. You can see the file creation process with another terminal, and can see the activity in the hosts’ disk performance view in the vSphere Client.
Once done, you’ll see it in the terminal and in the datastore view screen.
If you have a VMware or Hyper-V snapshot that has filled the drive, just hit the datastore browser and delete the failsafe file! You just gained time to better manage the snapshot or move data around to accommodate the growth.
I know this seems unnecessary, but every VMware admin seems to get burned with a rogue snapshot once. Occasionally, it’s more than once.
How can I prevent actually needing this?
That part is pretty simple, too. Just set up a vCenter alert to warn you if you have a snapshot that is growing out of control! The directions for how to do this are located in this VMware KB article. (Just remember to set up the email server so you actually receive the emails!)
This post is reposted from an older post but was updated for vSphere 6.5.