Yellow Bricks

by Duncan Epping

esa

vSAN 8.0 U1 – Disaggregated Storage Enhancements!

· · 2 Comments

With vSAN 8.0 U1 a lot of new features and enhancements are introduced. There are many blog posts out there describing the long list of enhancements, but in this post, I want to focus on HCI Mesh or Disaggregated vSAN specifically. (Also read this post by Cato!) For this feature, which in the UI is referred to as “Datastore Sharing”, there are 3 key enhancements introduced in vSAN 8.0 U1. There are enhancements for both the Original Storage Architecture (OSA), as well as the Express Storage Architecture (ESA).

With vSAN 8.0 the initial version of ESA was launched, and it did not support the use of Datastore Sharing. Starting with vSAN 8.0 U1 though, vSAN ESA is now also capable of sharing its storage with other clusters in the environment. To be more precise, a vSAN ESA cluster can now mount the datastore of another vSAN ESA cluster. What we also support is a “compute only” cluster mounting the vSAN ESA datastore remotely. So for those planning on implementing vSAN ESA, I think that is a very welcome enhancement!

For OSA there are also two enhancements for Datastore Sharing. The first I want to discuss is cross-vCenter Server datastore sharing. This feature is especially useful with customers who have a larger estate and are managing multiple clusters via different vCenter Server instances. You simply now have the option to connect the vCenter Server instances from a storage point of view, and then you can simply select the remote datastore in the cluster managed by a different vCenter Server instance. Let me just show you how this actually works in the next demo.

The second enhancement for OSA specifically is support for Stretched Cluster configurations. Starting with vSAN 8.0 U1 it is now possible to mount a vSAN Datastore which is stretched across locations. Your “client” cluster” can be “stretched”, “standard”, or compute-only even. We support all of those combinations. On top of that, the interface enables you to specify which location should be paired with which location, or fault domain. In other words, if you look at the diagram below, I can ensure that the hosts in Site A connect via the “local” network” to the remote datastore as part of Site A. This avoids IO traversing the intersite link, which can make a big difference in terms of latency and available bandwidth for other I/O etc.

I can imagine that the concepts are difficult to grasp without seeing the vSphere Client, so I spend some time in the lab to create a demo for you that walks you through the steps of how to configure this. In the lab I created a vSAN Stretched Cluster, and a standard cluster, and I am going to mount the vSAN stretched Datastore to the host in the standard cluster. Enjoy!

vSAN ESA ReadyNode configurations are more flexible than you think!

· · Leave a Comment

I had a discussion at the Dutch VMUG yesterday about the ReadyNode configurations for vSAN ESA. The discussion was about how difficult it was to select a host and customize it. It was then that I realized that most people hadn’t noticed yet that there is an easier method (or lifehack as my kids would say) when it comes to selecting your server model. How does that work? Well, let me show you!

First, let’s take a look at the vSAN ESA ReadyNode Hardware Guidance Table. The table below shows you what the node capacity is for each profile from a storage, CPU, memory, and networking perspective.

Now if you look at the table you will see that as the “profile” number goes up, so does the capacity for each of the various components. This is actually what provides you with a lot of flexibility in my opinion. If we take Dell as an example, but the same applies for most vendors on the current list, and we select “vSAN-ESA-AF2” and look at the list of options we see the following:

  • PowerEdge R650
  • PowerEdge R6515
  • PowerEdge R750
  • PowerEdge R7515

Now, if we look at “vSAN-ESA-AF8” next, which is the highest profile, we see that we only can pick 1 server model, which happens to be the PowerEdge R750. If we then look at the difference between the hosts selected for each profile a few things stand out:

  • vSAN-ESA-AF2 has an Intel Xeon Silver 4314, while vSAN-ESA-AF8 has a Platinum 8358
  • vSAN-ESA-AF2 has 512GB, while vSAN-ESA-AF8 has 1024GB
  • vSAN-ESA-AF2 a 25Gbps NIC, while vSAN-ESA-AF8 has a 100Gbps NIC
  • vSAN-ESA-AF2 has five 3.2TB NVMe devices while vSAN-ESA-AF8 has twenty-four 3.2TB devices

Now if I look at the KB article which explains what you can, and cannot change, something stands out, most of the components can be modified/customized. For instance, for CPU you can go to a higher core count and/or higher base clock speed! For memory, you can go up, same for storage devices (as long as you stay within supported limits), etc etc.

In other words, what is the difference between a vSAN-ESA-AF2 and a vSAN-ESA-AF8? Basically the expected workload, the performance, the capacity. This ultimately results in a different configuration. Nothing, at this point in time, stops you from selecting the “lowest” vSAN ReadyNode Profile and spec it as an “AF4”, “AF6” or “AF8” from a CPU stance, or from a storage/memory capacity point of view. If you want to have some more flexibility, try selecting a smaller profile, select the host type, and increase the resources/components where needed!

When you start exploring the options it may seem complex, but when you look more closely you will quickly realize that it actually isn’t that complex, and that it actually provides you with a lot of flexibility, as long as you stick to the rules and pick supported components!

vSAN ESA is using more CPU cycles than vSAN OSA?

· · 6 Comments

Over the last couple of weeks, I’ve had conversations with customers and partners who have been running performance benchmarks against both vSAN ESA and vSAN OSA. As you can imagine, people want to compare version 8 of OSA against version 8 of ESA, and that is completely fair. What I noticed though is that some of those customers came back with comments around CPU usage of vSAN OSA against ESA. The general comment we get is that vSAN ESA is using more CPU cycles than vSAN OSA.

When looking at it from a total number point of view, or CPU cycles consumed, it is very likely you will see vSAN ESA using more cycles than vSAN OSA. The question then typically arises why that is the case, as VMware (the vSAN team) has been claiming that vSAN ESA is much more efficient than vSAN OSA. To be fair, it is much more efficient. For instance data services like checksumming, encryption, and compression have moved to the top of the stack (as shown below) resulting in the fact that we don’t have to compress/encrypt data 3/4/5/6 times but can do it once at the source and then send it over the network to the destination.

Still, it leaves the question, why is more CPU capacity used? The answer is simple, you are pushing much more IO. We’ve seen customers easily reaching 4x the number of IOPS with ESA than with OSA. Even though ESA is more efficient, if you are pushing 4x (or more) the amount of IO then you will need to remember that those additional IOs also come at a cost, and that cost is CPU cycles to process them. So when you make a comparison, please compare apples to apples, and not apples to oranges.

The last thing I want to add, and hopefully I can share some data in the future, the use of RDMA with vSAN 8 ESA seems to have a significant impact on CPU usage, as in lower the amount of CPU required to produce the same results (or better results). So it is worth considering RDMA for sure when adopting vSAN 8 ESA!

What can I change about a vSAN ESA Ready Node?

· · Leave a Comment

I’ve had half a dozen people asking about this over the past weeks, it really seems more and more people are at the point of adopting vSAN ESA (Express Storage Architecture. When they look at the various vSAN ESA Ready Node configurations what stands out is that the current list is limited in terms of server models and configurations. (https://vmwa.re/vsanesahcl)

The list is being updated every week, last week for instance Supermicro popped up as a Server vendor. Of course, Dell, HPe, and Lenovo had been on the list since day 1. When you select the vendor, the ready node type, and the model you will then have the option to select a number of things, but in most cases, you seem to be limited to “Storage Device” and “Number of Storage Devices”. This however does not mean you cannot change anything. A knowledge base article has been released which describes what you can, and cannot change when it comes to these configurations! The KB article is listed on the vSAN ESA VMware Compatibility Guide list, but somehow it seems people don’t always notice the link. (Yes, I have asked the team to make the link more obvious somehow.)

Now when you look at the KB it lists what you can change, and what the rules are when it comes to making changes. For instance, you can change the CPU, but only for the same or higher core count and the same or higher base clock speed. For memory, you can increase the amount, and the same applies to storage capacity for instance. For storage it is even a bit more specific, you need to use the same make/model, so basically if the ReadyNode configuration lists a P5600 of 1.6TB, you can swap it for a P5600 of 3.2TB. However, that 3.2TB device will need to be certified for vSAN ESA as well, you can validate that using the following link: http://vmwa.re/vsanesahclc. Anyway, if you are configuring a Ready Node for ESA, make sure to check the KB so that you make supported changes!

Should I always use vSAN ESA, or can I go for vSAN OSA?

· · 4 Comments

Starting to get this question more often, should I always use vsAN ESA, or are there reasons to go for vSAN OSA? The answer is simple, whether you should use ESA or OSA can only be answered by the most commonly used phrase in consultancy: it depends. What does it depend on? Well, your requirements and your constraints.

One thing which comes up frequently is the 25Gbps requirement from a networking perspective. I’ve seen multiple people saying that they want to use ESA but their environment is not even close to saturating 10Gbps currently, so can they use ESA with 10Gbps? No, you cannot. ESA requires 25Gbps at minimum, and the bandwidth requirement fully depends on the ESA ReadyNode configuration you select. Why? Well, with ESA there’s also a certain performance expectation, which is why there’s a bandwidth requirement. The bandwidth requirement is put in place to ensure that you can use the NVMe devices to their full potential.

With both ESA and OSA you can produce impressive performance results, the big difference between ESA and OSA is the fact that ESA does this with a single type of NVMe device across the cluster, whereas OSA uses caching devices and capacity devices. ESA has also been optimized for high performance and is better at leveraging the existing host resources to achieve those higher numbers. An example of how ESA achieves that is through multi-threading for instance. What I appreciate about ESA the most is that the stack is also optimized in terms of resource usage. By moving data services to the top of the stack, data processing (compression of encryption for example) happens at the source instead of at bottom/destination. In other words, blocks are compressed by one host, and not by two or more (depending on the selected RAID level and type). Also, data is transferred over the network compressed, which saves bandwidth etc.

Back to OSA, why would you go for the Original Storage Architecture instead of ESA? Well, like I said, if you don’t have the performance requirements that dictate the use of ESA. If you want to use vSAN File Services (not supported with ESA in 8.o), HCI Mesh etc. If you want to run a hybrid configuration. If you want to use the vSAN Standard license. Plenty of reasons to still use OSA, so please don’t assume ESA is the only option. Use what you need to achieve your desired outcome, use what fits your budget, and use what will work with your constraints and requirements.