Explanation: Storage I/O Control v2Storage I/O Control (SIOC) was initially introduced in vSphere 4.1 to provide I/O prioritization of virtual machines running on a cluster of ESXi hosts that had access to shared storage. It extended the familiar constructs of shares and limits, which existed for CPU and memory, to address storage utilization through a dynamic allocation of I/O queue slots across a cluster of ESXi servers. The purpose of SIOC is to address the ‘noisy neighbor’ problem, i.e. a low priority virtual machine impacting other higher priority virtual machines due to the nature of the application and its I/O running in that low priority VM.
vSphere 5.0 extended SIOC to provide cluster-wide I/O shares and limits for NFS datastores. This means that no single virtual machine should be able to create a bottleneck in any environment regardless of the type of shared storage used. SIOC automatically throttles a virtual machine which is consuming a disparate amount of I/O bandwidth when the configured latency threshold has been exceeded. To allow other virtual machines receive their fair share of I/O bandwidth on the same datastore, a share based fairness mechanism has been created which now is supported on both NFS and VMFS.
vSphere 5.1 introduced a new SIOC feature called Stats Only Mode. When enabled, it doesn’t enforce throttling but gathers statistics to assist Storage DRS. Storage DRS now has statistics in advance for new datastores being added to the datastore cluster & can get up to speed on the datastores profile/capabilities much quicker than before.
Another 5.1 feature was Automatic Threshold Computation. The default latency threshold for SIOC is 30ms. Not all storage devices are created equal so this default was chosen as a sort of “catch-all”. There are certain devices which will hit their natural contention point much earlier than others, for example All Flash Arrays, in which case the threshold should be lowered by the user. However, manually determining the correct latency can be difficult for users. This gave rise to the need for the latency threshold to get automatically determined at a correct level for each device. Using the I/O injector modeling of SIOC, peak throughput and corresponding latency of a datastore is measured. The latency threshold value at which Storage I/O Control will kick in is then set to 90% of this peak value (by default). vSphere administrators can change this 90% to another percentage value or they can still input a millisecond value if they so wish.
The default latency threshold for SIOC can be reduced to as low as 5ms.
SIOC V1 OverviewSIOC V1 is disabled by default. It needs to be enabled on a per datastore level, and it is only utilized when a specific level of latency has been reached. By default, the latency threshold for a datastore is set to 30ms, as mentioned earlier. If SIOC is triggered, disk shares (aggregated from all VMDKs using the datastore) are used to assign I/O queue slots on a per host basis to that datastore. In other words, SIOC limits the number of IOs that a host can issue. The more VMs/VMDKs that run on a particular host, the higher the number of shares, and thus the higher the number of IOs that that particular host can issue. The throttling is done by modifying the device queue depth of the various hosts sharing the datastore. When the period of contention passes, and latency returns to normal values, the device queue depths are allowed to return to default values on each host.
SIOC V2 IntroductionBefore describing SIOC V2, it should be highlighted that SIOC V1 and SIOC V2 can co-exist on vSphere 6.5. This makes it much simpler when considering upgrades, or migrations between versions. With that in mind, SIOC V2 is considerably different from a user experience perspective when compared to V1. SIOCv2 is implemented using IO Filter framework Storage IO Control category. SIOC V2 can be managed using SPBM Policies. What this means is that you create a policy which contains your SIOC specifications, and these policies are then attached to virtual machines.
Creating an SIOC policy basedCreating an SIOC policy is done is exactly the same way as building a storage policy for VSAN or Virtual Volumes. Select the VM Storage Policy from the vSphere client home page, and from there select the option to create a new VM Storage Policy. VM Storage Policies in vSphere 6.5 has a new option called “Common Rules”. These are used for configuring data services provided by hosts, such as Storage I/O Control and Encryption.
Use common rules in the VM storage policyThe first step is to click on the check box to enabled common rules. This will then allow you to add components, such as SIOC, to the policy.
Add Component – Storage I/O ControlIn vSphere 6.5, there are two components available for common rules, Encryption and Storage I/O Control. Select Storage I/O Control in this case. Now you can select Normal, High, Low or Custom shares allocation.
This table describes the different Limits,Shares and Reservations associated with each setting:
HIGH
NORMAL
LOW
Limits
100,000
10,000
1,000
Reservation
100
50
10
Shares
2,000
1,000
500
When the policy has been created, it may be assigned to newly deployed VMs during provisioning,or to already existing VMs by assigning this new policy to the whole VM (or just an individual VMDK) by editing its settings. One thing to note is that IO Filter based IOPS does not look at the size of the IO. For example, there is no normalization so that a 64K IOP is not equal to 2 x 32K IOPS. It is a fixed value of IOPS irrespective of the size of the IO.
Custom AllocationIf neither of the values in the Normal, High, Low allocations is appropriate, there is the ability to create custom settings for these values. In a custom setting, IOPS limit and IOPS reservation are both set to -1, implying unlimited. These may be modified as required.
Advanced OptionsSchedCostUnitThis is an advanced parameter that was created for SIOC V1 only. SIOC V2 does not have SchedCostUnit implemented. For V1, SchedCostUnit determines the unit size (normalized size) of an IO operation for scheduling, and it is currently a constant value of 32K. This constant value, however, may not satisfy different requirements from different customers. Some customers may want to set this unit size to 4K. Other customers may want to set it up to 256K.
To satisfy these different requirements, SchedCostUnit is now configurable. It defaults to an IO size value of 32K, and allowable values range between 4K to 256K.
The SchedCostUnit dictates how requests are counted. A request with size <= SchedCostUnit counts as a single I/O. Anything greater than SchedCostUnit will be counted as 2 or more requests.
For example, by changing the SchedCostUnit from 32K to 64K, the number of IOPS observed will halve. The size of the IO can be set using the:
“esxcli system settings advanced set -o /Disk/SchedCostUnit -i 65536”
and verified by using the”
"esxcli system settings advanced list -o /Disk/SchedCostUnit”
command. SIOC V2 counts guest IO directly. IOPS will be counted based on IO count, regardless of the IO size.
SchedReservationBurstWhen limits are set on VMDKs, requests could have high average latency because the limit was enforced at a high (per request) granularity. This was due to the strict enforcement on a VM getting its share of IOs in interval of 1 second/L, where L is the user specified limit. The issue is more visible in fast storage, such as flash arrays. It was noted that SIOC V2 did not perform well when presented with a “bursty” workload on fast storage.
This SchedReservationBurst setting relaxes that constraint so a VM get its share of IOs at any time during a 1 second window, rather than enforce strict placement of IOs in intervals of 1/L. BURST option is turned-on by default.
SIOC V2 LimitationsIn this initial release of SIOC V2 in vSphere 6.5, there is no support for vSAN or Virtual Volumes. SIOC v2 is only supported with VMs that run on VMFS and NFS datastores.
