Tag Archives: IOPS

Full NVMe or not Full NVMe, that is the question

Image result for nvme logo

As you have seen, my recent posts have been around Intel Optane and the performance gains that can be delivered by implementing the technology into a vSAN environment. I have been asked many times about what benefits a full NVMe solution would bring and what such a solution would look like, but before we go into that, let’s talk about NVMe, what exactly is NVMe?

Non-Volatile Memory Express (NVMe) is not a drive type, but more of an interface and protocol solution that looks like is set to replace the SAS/SATA interface. It encompasses a PCIe controller and the whole purpose of NVMe is to exploit the parallelism that flash media provides which in turn reduces the I/O overhead and thus improve performance. As SSDs become faster, protocols like SAS/SATA which were designed for slower hard disks where the delay between the CPU request and data transfer was much higher, the requirement for faster protocols become evident, and this is where NVMe comes into play.

So in a vSAN environment, what does a full NVMe solution look like? Because vSAN is currently a two tier architecture (Cache and Capacity) a full NVMe solution would mean that both tiers have to have NVMe capable drives and this can be done with either all Standard NVMe drives in both cache and capacity, or using a technology like Intel Optane NVMe as the Cache and Standard NVMe as capacity. So from an architecture perspective it is pretty straight forward, but how does performance compare, for this I persuaded my contacts at Intel to provide me some Full NVMe kit in order to perform some benchmark tests, and in order to provide a like for like comparison, I ran the same benchmark tests on an Optane+SATA configuration.

Cluster Specification:
Number of Nodes: 4
Network: 2x 10gbit in LACP configuration
Disk groups per node: 2
Cache Tier both clusters: 2x Intel Optane 375GB P4800X PCIe Add In Card
Capacity Tier Optane/SATA: 8x 3.84TB SATA S4510 2.5″
Capacity Tier Full NVMe: 8x 2.0TB NVMe P4510 2.5″ U.2

Test Plan:
Block Size: 4K, 8K, 16K, 32K, 64K, 128K
I/O Pattern: Random
Read/Write Ratio: 0/100, 30/70, 70/30, 100/0
Number of VMs: 120
Number of VMDKs per VM: 1
Size of VMDK: 50GB
Storage Policy: FTT=1, RAID1

Let’s look at the results:

And if you want the numbers:

So what is clear here that Optane serves really well in the cache tier in both solutions, however in the Full NVMe solution read performance is significantly improved also, in the 128K, 100% read test the 2x10G Links were being pushed to their limits, but not only was we able to push up throughput and IOPS but we also drove down latency, in some cases reducing it by over 50%.

So why would you choose a full NVMe solution? The simple answer here is if you have applications that are latency sensitive then having clusters dedicated to those applications would be adequately provided for from an IOPS, Throughput and Latency perspective with Full NVMe.

Vendors have also recognised this, for example Dell EMC have just launched their Intel Optane Powered Full NVMe vSAN Ready node, based on the R740xd platform and consists of similar drives to what I have used in the tests here being the Optane 375GB and P4510 U.2 NVMe drives, you can see the vSAN ready node details here

So clearly NVMe has major performance benefits over traditional SAS/SATA devices, could this be the end of SAS/SATA in the not so distant future?

Optane Performance

Many times over the past few months I have been asked about the benefits of using Intel Optane NVMe in a vSAN environment, although there was marketing material from Intel that boasted a good performance boost I decided (purely out of curiosity) to do some performance benchmarking and compare Optane as the cache devices versus SAS as the cache devices. The performance benchmark test used exactly the same servers and networking in order to provide a level playing field, the only thing that was changed was the cache devices being used in the disk groups.

Server Specification:

  • 6x Dell PowerEdge R730xd
  • Intel Xeon CPU E5-2630 v3 @ 2.40GHz
  • 128GB RAM
  • 2x Dell PERC H730 Controllers
  • 2x Intel Dual Port 10Gb ethernet adapters (Configured with LACP)

Disk group config for the SAS test:

  • 3x Disk Groups
  • 3x 400GB SAS SSD per disk group
  • 1x 400GB SAS SSD per disk group

Disk group config for the Optane test

  • 2x Disk Groups
  • 3x 400GB SAS SSD per disk group
  • 1x 750GB Optane NVMe P4800X per disk group

Whilst you could say that the configurations are not identical, since the Write Buffer is limited to 600GB per disk group then both configurations have the same amount of write buffer, the SAS config has more backend disks which would serve as an advantage.

For the purpose of the Benchmark, we used HCI Bench to automate the Oracle VDBench workload testing and each test was based on the following, the test was designed to max-out the system hence the high number of VMDKs used here (250)

  • 50 Virtual Machines
  • 5 VMDKs per virtual machine
  • 2 threads per VMDK
  • 20% working set
  • 4k, 8k, 16k, 32k, 64k and 128k block size
  • 0%, 30%, 70%, 100% write workload
  • 900 second test time for each test

So what were the results?

4K Blocksize:

8K Blocksize:

16K Blocksize:

32K Blocksize:

64K Blocksize:

128K Blocksize:

As you can see Optane really did boost the performance even though the server platform wasn’t the ideal platform for the Optane devices (Dell said those cards will not be certified in the 13G platform), however despite the fact that the workload was designed to max-out the system, in some cases latency was reduced to almost a third and throughput was was increased in some cases to 3x.

Conclusion: Optane really does live up to expectations, and it isn’t just marketing, I have yet to test a full NVMe system to see how much it can really be pushed, but I hope the numbers above go someway to convice you why you should consider optane as the cache tier in vSAN.

Day-2 Operations – vSphere built in vROPS dashboards

At VMworld I ran a few sessions on Day-2 Operations which I also covered the new built in dashboards for vROPS which were introduced with the fully baked HTML5 client in vSAN/vSphere 6.7.  Many people were not aware of the dashboards and moreso were not aware that these vSAN specific dashboards continue to work even after the 90 day trial period has expired.  Not only that but VMware has done a great job in automating a deployment of vROPS if you have not already got an appliance deployed.  So let’s take a look at these new dashboards in a bit more detail.

Firstly the three vSphere operations Dashboards, the default one that loads up is an overview dashboard of all your clusters:

Then there is the Cluster Level View where you pick a specific cluster:

And finally is the Alerts view…Great timing, I have a physical disk failure in the vSAN Cluster:Now the above three dashboards will no longer be available after the 90 day trial period has expired, and also the link to the vRealize Operations appliance will not be functional either, but after the 90 day trial period has expired, the following three vSAN Dashboards will still be fully functional and available, so let’s look at those in a bit more detail:

The vSAN Overview dashboard, like the vSphere overview dashboard, displays information at a holistic level for all of your vSAN Clusters within this particular vCenter server, you will see that the dashboard provides information on how many clusters are running dedupe/compression, or how many of the clusters are a Stretched Cluster for example.  The dashboard also shows if you need to investigate any current alerts (yes I cleaned up the failed disk before grabbing this screenshot).

The next dashboard we choose a specific cluster to look at in more detail:
In this dashboard we see information pertaining to a specific cluster, we can see that I have 6 critical alerts which we will take a look at next, but there are some key metrics here that from an operational perspective are pretty important from a day-2 operations standpoint:

  • Remaining Capacity
  • Component Limit
  • IOPS, Throughput and Latency statistics
  • Read versus Write latency

The last dashboard is alerts:
Here we can see the current alerts that have been triggered for each cluster which may need to be addressed, the critical alerts previously highlighted in the cluster view were all related to a network redundancy lost when I was troubleshooting packet loss on the physical switch.

So as you can see there’s a really good amount of detail in the vSphere UI relating to vROPS making the day-2 operations a lot easier to perform.