SIOC Archives - frankdenneman.nl

How to enable SIOC stats only mode?

February 11, 2013 by frankdenneman 5 Comments

Today on twitter, David Chadwick, Cormac Hogan and I were discussing SIOC stats only mode. SIOC stats only mode gathers statistics to provide you insights on the I/O utilization of the datastore. Please note that Stats only mode does not enable the datastore-wide scheduler and will not enforce throttling. Stats only mode is disabled due to the (significant) increase of log data into the vCenter database.
SIOC stats only mode is available from vSphere 5.1 and can be enabled via the web client. To enable SIOC stats only mode go to:

Storage view
Select the datastore
Select Manage
Select Settings

By default both SIOC and SIOC stats only mode is disabled. Click on the edit button at the right side of the screen. Un-tick the check box “Disable Storage I/O statistics collection (applicable only if Storage I/O Control is disabled)”. Click on OK

To test to see if there is any difference, I used a datastore that SIOC had enabled. I disabled SIOC and un-ticked the “Disable Storage I/O statistics collection (applicable only if Storage I/O Control is disabled)” option. I opened up the performance view and selected the “Realtime” Time Range.

Storage view
Select the datastore
Select Monitor
Select Performance
Select “Realtime” Time range

At 15:35 I disabled SIOC, which explains the dip, at 15:36 SIOC stats only mode was enabled and it took vCenter roughly a minute to start displaying the stats again.

As all new vSphere 5.1 features, SIOC stats only mode can only be enabled via the vSphere web client.

SIOC on datastores backed by a single datapool

December 6, 2012 by frankdenneman 5 Comments

Duncan posted an article today in which he brings up the question: Should I use many small LUNs or a couple large LUNs for Storage DRS? In this article he explains the differences between Storage I/O Control (SIOC) and Storage DRS and why they work well together, to re-emphasize, the goal of Storage DRS load balancing is to fix long term I/O imbalances, while SIOC addresses short term burst and loads. SIOC is all about managing the queue’s while Storage DRS is all about intelligent placement and avoiding bottlenecks.
Julian Wood makes an interesting remark, and both Duncan and I hear this remark when discussing SIOC. Don’t get me wrong I’m not picking on Julian, I’m merely stating the fact he made a frequently used argument.

“There is far less benefit in using Storage IO Control to load balance IO across LUNs ultimately backed by the same physical disks than load balancing across separate physical storage pools. “

Well when you look at the way SIOC works I tend to disagree with this statement. As stated before, SIOC manages queues, queues to the datastores used by the virtual machines in the virtual datacenter. Typically speaking these virtual machines differ from workload types, from peak moments and also they differ in importance to the organization. With the use of disk shares, important virtual machine can be assigned a higher priority within the disk queue. When contention occurs, and this is important to realize, when contention occurs these business critical virtual machine get prioritized over other virtual machines. Not all important virtual machines generate a constant stream of I/O, while other virtual machines, maybe with a lower priority do generate a constant stream of IO. The disk shares provide the high priority low IO virtual machines to get a foot between the door and get those I/Os to the datastore and back. Without SIOC and disk shares you need to start thinking of increasing the queue depth of each hosts and think about smart placement of these virtual machines (both high and low I/O load) to avoid those high I/O load getting on the same host. These placement adjustment might impact DRS load balancing operations, possibly affecting other virtual machines along the way. Investing time in creating and managing a matrix of possible vm to datastore placement is not the way to go in this time with rapidly expanding datacenters.
Because SIOC is a datastore-wide scheduler, SIOC determines the queue-depth of the ESX hosts connected to the datastores running virtual machines on those datastores. Hosts with higher priority virtual machines get “deeper” queue depths to the datastore and hosts with lower priority virtual machines running on the datastore receive shorter queue-depths. To be more precise, SIOC calculates the datastore wide latency and each local host scheduler determines the queue depth for the queues of the datastore.
But remember queue depth changes only occur when there is contention, when the datastore exceeds the SIOC latency threshold. For more info about SIOC latency read “To which Host level latency statistic is the SIOC threshold related”
Coming back to the argument, I firmly believe that SIOC has benefits in a shared diskpool structure, between the VMM and the datastore a lot of queue’s exists.

Because SIOC takes the avg device latency off all hosts connected to the datastore into account, it understands the overall picture when determining the correct queue depth for the virtual machines. Keep in mind, queue depth changes occur only during contention. Now the best part of SIOC in 5.1 is that it has the Automatic Latency Threshold Computation. By leveraging the SIOC injector it understands the peak value of a datastore and adjust the SIOC threshold. The SIOC threshold will be set to 90% of its peak value, therefor having an excellent understanding of the performance capability of the datastore. This is done on a regular basis so it keeps actual workload in mind. This dynamic system will give you far more performance benefit that statically setting the queue-depth and DNSRO for each host.
One of the main reasons of creating multiple datastores that are backed by a single datapool is because of creating a multi-path environment. Together with advanced multi-pathing policies and LUN to controller port mappings, you can get the most out of your storage subsystem. With SIOC, you can manage your queue depths dynamically and automatically, by understanding actually performance levels, while having the ability to prioritize on virtual machine level.

vSphere 5.1 Storage DRS load balancing and SIOC threshold enhancements

October 19, 2012 by frankdenneman 4 Comments

Lately I have been receiving questions on best practices and considerations for aligning the Storage DRS latency and Storage IO Control (SIOC) latency, how they are correlated and how to configure them to work optimally together. Let’s start with identifying the purpose of each setting, review the enhancements vSphere 5.1 has introduced and discover the impact when misaligning both thresholds in vSphere 5.0.
Purpose of the SIOC threshold
The main goal of the SIOC latency is to give fair access to the datastores, throttling virtual machine outstanding I/O to the datastores across multiple hosts to keep the measured latency below the threshold.

It can have a restrictive effect on the I/O flow of virtual machines.
Purpose of the Storage DRS latency threshold
The Storage DRS latency is a threshold to trigger virtual machine migrations. To be more precise, if the average latency (VMObservedLatency) of the virtual machines on a particular datastore is higher than the Storage DRS threshold, then Storage DRS will mark that datastore as “source” for load balancing migrations. In other words, that datastore provide the candidates (virtual machines) for Storage DRS to move around to solve the imbalance of the datastore cluster.

This means that the Storage DRS threshold metric has no “restrictive” access limitations. It does not limit the ability of the virtual machines to send I/O to the datastore. It is just an indicator for Storage DRS which datastore to pick for load balance operations.
SIOC throttling behavior
When the average device latency detected by SIOC is above the threshold, SIOC throttles the outstanding IO of the virtual machines on the hosts connected to that datastore. However due to different number of shares, various IO sizes, random versus sequential workload and the spatial locality of the changed blocks on the array, we are almost certain that no virtual machine will experience the same performance. Some virtual machines will experience a higher latency than other virtual machines running on that datastore. Remember SIOC is driven by shares, not reservations, we cannot guarantee IO slots (reservations). Long story short, when the datastore is experiencing latency, the VMkernel manages the outbound queue, resulting in creating a buildup of I/O somewhere higher up in the stack. As the SIOC latency threshold is the weighted average of D/AVG per host, the weight is the number of IOPS on that host. For more information how SIOC calculates the Device average latency, please read the article: “To which host-level latency statistic is the SIOC congestion threshold related?”
Has SIOC throttling any effect on Storage DRS load balancing?
Depending on which vSphere version you run is the key whether SIOC throttling has impact on Storage DRS load balancing. As stated in the previous paragraph, if SIOC throttles the queues, the virtual machine I/O does not disappear, vSphere always allows the virtual machine to generate I/O to the datastore, it just builds up somewhere higher in the stack between the virtual machine and the HBA queue.
In vSphere 5.0, Storage DRS measures latency by averaging the device latency of the hosts running VMs on that datastore. This is almost the same metric as the SIOC latency. This means that when you set the SIOC latency equal to the Storage DRS latency, the latency will be build up in the stack above the Storage DRS measure point. This means that in worst-case scenario, SIOC throttles the I/O, keeping it above the measure point of Storage DRS, which in turn makes the latency invisible to Storage DRS and therefore does not trigger the load balance operation for that datastore.
Introducing vSphere 5.1 VMObservedLatency
To avoid this scenario Storage DRS in vSphere 5.1 is using the metric VMObservedLatency. This metric measures the round-trip of I/O from the moment the VMkernel receives the I/O (Virtual Machine Monitor) to the datastore and all the way back to the VMM.

This means that when you set the SIOC latency to a lower threshold than the Storage DRS latency, Storage DRS still observes the latency build up in the kernel layer.
vSphere 5.1 Automatic latency SIOC
To help you avoid building up I/O in the host queue, vSphere 5.1 offers automatic threshold computation for SIOC. SIOC sets the latency to 90% of the throughput level of the device. To determine this, SIOC derives a latency setting after a series of tests, mapping maximum throughput to a latency value. During the tests SIOC detects where the throughput of I/O levels out, while the latency keeps on increasing. To be conservative, SIOC derives a latency value that allows the host to generate up to 90% of the throughput, leaving a burst space of 10%. This provides the best performance of the devices, avoiding unnecessary restrictions by building up latency in the queues. In my opinion, this feature alone warrants the upgrade to vSphere 5.1.
How to set the two thresholds to work optimally together
SIOC in vSphere 5.1 allows the host to go up to 90% of the throughput before adjusting the queue length of each host, and generate queuing in the kernel instead of queuing on the storage array. As Storage DRS uses VMObservedLatency it monitors the complete stack. It observes the overall latency, disregarding the location of the latency in the stack and tries to move VMs to other datastores to level out the overall experienced latency in the datastore cluster. Therefore you do not need to worry about misaligning the SIOC latency and the Storage DRS I/O latency.
If you are running vSphere 5.0 it’s recommended setting the SIOC threshold to a higher value than the Storage DRS I/O latency threshold. Please refer to your storage vendor to receive the accurate SIOC latency threshold.
Get notification of these blogs postings and more DRS and Storage DRS information by following me on Twitter: @frankdenneman

To which host-level latency statistic is the SIOC congestion threshold related?

July 23, 2012 by frankdenneman 10 Comments

Today someone asked if the congestion threshold of SIOC is related to which host latency threshold? Is it the Device average (DAVG), Kernel Average (KAVG) or Guest Average (GAVG)?

Well actually it’s none of the above. DAVG, KAVG and GAVG are metrics in a host-local centralized scheduler that has complete control over all the requests to the storage system. SIOC main purpose is to manage shared storage resources across ESXi hosts, providing allocation of I/O resources independent of the placement of virtual machines accessing the shared datastore. And because it needs to regulate and prioritize access to shared storage that spans multiple ESXi hosts, the congestion threshold is not measured against a host-side latency metric. But to which metric is it compared? In essence the congestion threshold is compared with the weighted average of D/AVG per host, the weight is the number of IOPS on that host. Let’s expand on this a bit further.
Average I/O latency
To have an indication of the load of the datastore on the array, SIOC uses the average I/O latency detected by each host connected to that datastore. Average latency across hosts is used to cope with the variety of workloads, the characteristic of the active workloads, such as read versus writes, I/O size and degree of sequential I/Os in addition to array behavior such as block location, caching policies and I/O scheduling.
To calculate and normalize the average latency across hosts, each host writes its average device latency and number of I/Os for that datastore in a file called IORMSTATS.SF stored on the same datastore.

A common misconception about SIOC is that it’s compute cluster based. The process of determining the datastore-wide average latency really reveals the key denominator – hosts connected to the datastore – . All hosts connected to the datastore write to the IORMSTATS.SF file, regardless of cluster membership. Other than enabling SIOC, vCenter is not necessary for normal operations. Each connected host reads the IORMSTATS.SF file each 4 seconds and locally computes the datastore-wide average to use for managing the I/O stream. Therefor cluster membership is irrelevant.
Datastore wide normalized I/O latency
Back to the process of computing the datastore wide normalized I/O latency. The average device latencies of each host are normalized by SIOC based on the I/O request size. As mentioned before, not all storage related workloads are the same. Workloads issuing I/Os with a large request size result in longer device latencies due to way storage arrays process these workloads. For example, when using a larger I/O request size such as 256KB, the transfer might be broken up by the storage subsystem into multiple 64KB blocks. This operation can lead to a decline of transfer rate and throughput levels, increasing latency. This allows SIOC to differentiate high device latency from actual I/O congestion at the device itself.
Number of I/O requests complete per second
At this point SIOC has normalized the average latency across hosts based on I/O size, next step is to determine the aggregate number of IOPS accessing the datastore. As each host reports the number of I/O requests complete per second, this metric is used to compare and prioritize the workloads.
I hope this mini-deepdive into the congestion thresholds explains why the congestion threshold could never be solely related to a single host-side metric . Because the datastore-wide average latency is a normalized value, the latency observed of the datastore per individual host may be different than the latency SIOC reports per datastore.
.