Welcome!

@CloudExpo Authors: Ed Featherston, Rostyslav Demush, Jamie Madison, Jason Bloomberg, Greg Pierce

Related Topics: @CloudExpo, Containers Expo Blog, Agile Computing

@CloudExpo: Blog Feed Post

Part 2: Some Server Storage I/O Workload Scripts and Results | @CloudExpo #Cloud #Storage #DataCenter

Everything is not the same in the data center or with data infrastructures that support different applications

Part 2: Some Server Storage I/O Workload Scripts and Results

This is the second in a two part series of posts pertaining to using some common server storage I/O benchmark tools and workload scripts. View part I here which includes overview, background and information about the tools used and related topics.

Following are some server I/O benchmark workload scripts to exercise various storage devices such as Non-Volatile Memory (NVM) flash Solid State Devices (SSDs) or Hard Disk Drives (HDD) among others.

NVMe ssd storage

Various NVM flash SSD including NVMe devices

The Workloads
Some ways that can impact the workload performance results besides changing the I/O size, read write, random sequential mix is the number of threads, workers and jobs. Note that in the workload steps, the larger 1MB and sequential scenarios have fewer threads, workers vs. the smaller IOP or activity focused workloads. Too many threads or workers can cause overhead and you will reach a point of diminishing return at some point. Likewise too few and you will not drive the system under test (SUT) or device under test (DUT) to its full potential. If you are not sure how many threads or workers to use, run some short calibration tests to see the results before doing a large, longer test.

Keep in mind that the best benchmark or workload is your own application running with similar load to what you would see in real world, along with applicable features, configuration and functionality enabled. The second best would be those that closely resemble your workload characteristics and that are relevant.

The following workloads involved a system test initiator (STI) server driving workload using the different tools as well as scripts shown. The STI sends the workload to a SUT or DUT that can be a single drive, card or multiple devices, storage system or appliance. Warning: The following workload tests does both reads and writes which can be destructive to your device under test. Exercise caution on the device and file name specified to avoid causing a problem that might result in you testing your backup / recovery process. Likewise no warranty is given, implied or made for these scripts or their use or results, they are simply supplied as is for your reference.

The four free tools that I'm using with this set of scripts are:

  • Microsoft Diskspd (free), get the tool and bits here or here (open source), learn more about Diskspd here.
  • FIO.exe (free), get the tool and bits here or here among other venues.
  • Vdbench (free with registration), get the tool and bits here or here among other venues.
  • Iometer (free), get the tool and bits here among other venues.

Notice: While best effort has been made to verify the above links, they may change over time and you are responsible for verifying the safety of links and your downloads

Microsoft Diskspd workloads
Note that a 300GB size file named iobw.tst on device N: is being used for performing read and write I/Os to. There are 160 threads, I/O size of 4KB and 8KB varying from 100% Read (0% write), 70% Read (30% write) and 0% Read (100% write) with random (seek) and no hardware or software cache. Also specified are to collect latency statistics, a 30 second warm up ramp up time, and a quick 5 minute duration (test time). 5 minutes is a quick test for calibration, verify your environment however relatively short for a real test which should be in the hours or more depending on your needs.

Note that the output results are put into a file with a name describing the test tool, workload and other useful information such as date and time. You may also want to specify a different directory where output files are placed.

diskspd.exe -c300G -o160 -t160 -b4K -w0 -W30 -d300 -h -fr  N:iobw.tst -L  > DiskSPD_300G_4KRan100Read_160x160_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b4K -w30 -W30 -d300 -h -fr  N:iobw.tst -L  > DiskSPD_300G_4KRan70Read_160x160_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b4K -w100 -W30 -d300 -h -fr  N:iobw.tst -L  > DiskSPD_300G_4KRan0Read_160x160_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b8K -w0 -W30 -d300 -h -fr  N:iobw.tst -L  > DiskSPD_300G_8KRan100Read_160x160_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b8K -w30 -W30 -d300 -h -fr  N:iobw.tst -L  > DiskSPD_300G_8KRan70Read_160x160_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b8K -w100 -W30 -d300 -h -fr  N:iobw.tst -L  > DiskSPD_300G_8KRan0Read_160x160_072416_8AM.txt

The following Diskspd tests use similar settings as above, however instead of random, sequential is specified, threads and outstanding I/Os are reduced while I/O size is set to 1MB, then 8KB, with 100% read and 100% write scenarios. The -t specifies the number of threads and -o number of outstanding I/Os per thread.

diskspd.exe -c300G -o32 -t132 -b1M -w0 -W30 -d300 -h -si  N:iobw.tst -L  > DiskSPD_300G_1MSeq100Read_32x32_072416_8AM.txt
diskspd.exe -c300G -o32 -t132 -b1M -w100 -W30 -d300 -h -si  N:iobw.tst -L  > DiskSPD_300G_1MSeq0Read_32x32_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b8K -w0 -W30 -d300 -h -si  N:iobw.tst -L  > DiskSPD_300G_8KSeq100Read_32x32_072416_8AM.txt
diskspd.exe -c300G -o160 -t160 -b8K -w100 -W30 -d300 -h -si  N:iobw.tst -L  > DiskSPD_300G_8KSeq0Read_32x32_072416_8AM.txt

Fio.exe workloads
Next are the fio workloads similar to those run using Diskspd except the sequential scenarios are skipped.

fio --filename=N\:\iobw.tst --filesize=300000M --direct=1  --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=windowsaio  --ba=4k --bs=4k --rwmixread=100 --iodepth=32 --numjobs=5 --exitall --time_based  --ramp_time=30 --runtime=300 --group_reporting --name=xxx  --output=FIO_300000M_4KRan100Read_5x32_072416_8AM.txt
fio --filename=N\:\iobw.tst --filesize=300000M --direct=1  --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=windowsaio  --ba=4k --bs=4k --rwmixread=70 --iodepth=32 --numjobs=5 --exitall --time_based  --ramp_time=30 --runtime=300 --group_reporting --name=xxx  --output=FIO_300000M_4KRan70Read_5x32_072416_8AM.txt
fio --filename=N\:\iobw.tst --filesize=300000M --direct=1  --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=windowsaio  --ba=4k --bs=4k --rwmixread=0 --iodepth=32 --numjobs=5 --exitall --time_based  --ramp_time=30 --runtime=300 --group_reporting --name=xxx  --output=FIO_300000M_4KRan0Read_5x32_072416_8AM.txt
fio --filename=N\:\iobw.tst --filesize=300000M --direct=1  --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=windowsaio  --ba=8k --bs=8k --rwmixread=100 --iodepth=32 --numjobs=5 --exitall --time_based  --ramp_time=30 --runtime=300 --group_reporting --name=xxx  --output=FIO_300000M_8KRan100Read_5x32_072416_8AM.txt
fio --filename=N\:\iobw.tst --filesize=300000M --direct=1  --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=windowsaio  --ba=8k --bs=8k --rwmixread=70 --iodepth=32 --numjobs=5 --exitall --time_based  --ramp_time=30 --runtime=300 --group_reporting --name=xxx  --output=FIO_300000M_8KRan70Read_5x32_072416_8AM.txt
fio --filename=N\:\iobw.tst --filesize=300000M --direct=1  --rw=randrw --refill_buffers --norandommap --randrepeat=0 --ioengine=windowsaio  --ba=8k --bs=8k --rwmixread=0 --iodepth=32 --numjobs=5 --exitall --time_based  --ramp_time=30 --runtime=300 --group_reporting --name=xxx  --output=FIO_300000M_8KRan0Read_5x32_072416_8AM.txt

Vdbench workloads
Next are the Vdbench workloads similar to those used with the Microsoft Diskspd scenarios. In addition to making sure Vdbench is installed and working, you will need to create a text file called seqrxx.txt containing the following:

hd=localhost,jvms=!jvmn
sd=sd1,lun=!drivename,openflags=directio,size=!dsize
wd=mix,sd=sd1
rd=!jobname,wd=mix,elapsed=!etime,interval=!itime,iorate=max,forthreads=(!tthreads),forxfersize=(!worktbd),forseekpct=(!workseek),forrdpct=(!workread),openflags=directio

The following are the commands that call the Vdbench script file. Note Vdbench puts output files (yes, plural there are many results) in a output folder.

vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=4k workseek=100 workread=100 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o  vdbench_NNVMe_300GB_64JVM_160TH_4K100Ran100Read_0726166AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=4k workseek=100 workread=70 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_4K100Ran70Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=4k workseek=100 workread=0 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_4K100Ran0Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=8k workseek=100 workread=100 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_8K100Ran100Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=8k workseek=100 workread=70 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_8K100Ran70Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=8k workseek=100 workread=0 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_8K100Seq0Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=8k workseek=0 workread=100 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_8K100Seq100Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=8k workseek=0 workread=70 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_8K100Seq70Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=160 jvmn=64 worktbd=8k workseek=0 workread=0 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_160TH_8K100Seq0Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=32 jvmn=64 worktbd=1M workseek=0 workread=100 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_32TH_1M100Seq100Read_072416_8AM
vdbench -f seqrxx.txt dsize=300G  tthreads=32 jvmn=64 worktbd=1M workseek=0 workread=0 jobname=NVME etime=300 itime=30 drivename="\\.\N:\iobw.tst" -o vdbench_NNVMe_300GB_64JVM_32TH_1M100Seq0Read_072416_8AM

Iometer workloads
Last however not least, lets do an Iometer run. The following command calls an Iometer input file (icf) that you can find here. In that file you will need to make a few changes including the name of the server where Iometer is running, description and device under test address. For example in the icf file change SIOSERVER to the name of the server where you will be running Iometer from. Also change the address for the DUT, for example N: to what ever address, drive, mount point you are using. Also update the description accordingly (e.g. "NVME" to "Your test example".

Here is the command line to run Iometer specifying an icf and where to put the results in a CSV file that can be imported into Excel or other tools.

iometer /c iometer_5work32q_intel_Profile.icf /r iometer_nvmetest_5work32q_072416_8AM.csv

server storage I/O SCM NVM SSD performance

What About The Results?
For context, the following results were run on a Lenovo TS140 (32GB RAM), single socket quad core (3.2GHz) Intel E3-1225 v3 with an Intel NVMe 750 PCIe AiC (Intel SSDPEDMW40). Out of the box Microsoft Windows NVMe drive and controller drivers were used (e.g. 6.3.9600.18203 and 6.3.9600.16421). Operating system is Windows 2012 R2 (bare metal) with NVMe PCIe card formatted with ReFS file system. Workload generator and benchmark driver tools included Microsoft Diskspd version 2.012, Fio.exe version 2.2.3, Vdbench 50403 and Iometer 1.1.0. Note that there are newer versions of the various workload generation tools.

Example results are located here.

Where to Learn More

What This All Means
Remember, everything is not the same in the data center or with data infrastructures that support different applications.

While some tools are more robust or better than others for different things, ultimately its usually not the tool that results in a bad benchmark or comparison, its the configuration or lack of including workload settings that are not relevant or applicable. The best benchmark, workload or simulation is your own application. Second best is one that closely resembles your application workload characteristics. A bad benchmark is one that has no relevance to your environment, application use scenario. Take and treat all benchmark or workload simulation results with a grain of salt as something to compare, contrast or make reference to in the proper context.

Ok, nuff said, for now...

Cheers
Gs

Greg Schulz - Author Cloud and Virtual Data Storage Networking (CRC Press), The Green and Virtual Data Center (CRC Press) and Resilient Storage Networks (Elsevier)
twitter @storageio

All Comments, (C) and (TM) belong to their owners/posters, Other content (C) Copyright 2006-2016 Server StorageIO(R) and UnlimitedIO All Rights Reserved

Read the original blog entry...

More Stories By Greg Schulz

Greg Schulz is founder of the Server and StorageIO (StorageIO) Group, an IT industry analyst and consultancy firm. Greg has worked with various server operating systems along with storage and networking software tools, hardware and services. Greg has worked as a programmer, systems administrator, disaster recovery consultant, and storage and capacity planner for various IT organizations. He has worked for various vendors before joining an industry analyst firm and later forming StorageIO.

In addition to his analyst and consulting research duties, Schulz has published over a thousand articles, tips, reports and white papers and is a sought after popular speaker at events around the world. Greg is also author of the books Resilient Storage Network (Elsevier) and The Green and Virtual Data Center (CRC). His blog is at www.storageioblog.com and he can also be found on twitter @storageio.

@CloudExpo Stories
DX World EXPO, LLC, a Lighthouse Point, Florida-based startup trade show producer and the creator of "DXWorldEXPO® - Digital Transformation Conference & Expo" has announced its executive management team. The team is headed by Levent Selamoglu, who has been named CEO. "Now is the time for a truly global DX event, to bring together the leading minds from the technology world in a conversation about Digital Transformation," he said in making the announcement.
"Space Monkey by Vivent Smart Home is a product that is a distributed cloud-based edge storage network. Vivent Smart Home, our parent company, is a smart home provider that places a lot of hard drives across homes in North America," explained JT Olds, Director of Engineering, and Brandon Crowfeather, Product Manager, at Vivint Smart Home, in this SYS-CON.tv interview at @ThingsExpo, held Oct 31 – Nov 2, 2017, at the Santa Clara Convention Center in Santa Clara, CA.
SYS-CON Events announced today that Conference Guru has been named “Media Sponsor” of the 22nd International Cloud Expo, which will take place on June 5-7, 2018, at the Javits Center in New York, NY. A valuable conference experience generates new contacts, sales leads, potential strategic partners and potential investors; helps gather competitive intelligence and even provides inspiration for new products and services. Conference Guru works with conference organizers to pass great deals to gre...
DevOps is under attack because developers don’t want to mess with infrastructure. They will happily own their code into production, but want to use platforms instead of raw automation. That’s changing the landscape that we understand as DevOps with both architecture concepts (CloudNative) and process redefinition (SRE). Rob Hirschfeld’s recent work in Kubernetes operations has led to the conclusion that containers and related platforms have changed the way we should be thinking about DevOps and...
In his Opening Keynote at 21st Cloud Expo, John Considine, General Manager of IBM Cloud Infrastructure, led attendees through the exciting evolution of the cloud. He looked at this major disruption from the perspective of technology, business models, and what this means for enterprises of all sizes. John Considine is General Manager of Cloud Infrastructure Services at IBM. In that role he is responsible for leading IBM’s public cloud infrastructure including strategy, development, and offering m...
The next XaaS is CICDaaS. Why? Because CICD saves developers a huge amount of time. CD is an especially great option for projects that require multiple and frequent contributions to be integrated. But… securing CICD best practices is an emerging, essential, yet little understood practice for DevOps teams and their Cloud Service Providers. The only way to get CICD to work in a highly secure environment takes collaboration, patience and persistence. Building CICD in the cloud requires rigorous ar...
Companies are harnessing data in ways we once associated with science fiction. Analysts have access to a plethora of visualization and reporting tools, but considering the vast amount of data businesses collect and limitations of CPUs, end users are forced to design their structures and systems with limitations. Until now. As the cloud toolkit to analyze data has evolved, GPUs have stepped in to massively parallel SQL, visualization and machine learning.
"Evatronix provides design services to companies that need to integrate the IoT technology in their products but they don't necessarily have the expertise, knowledge and design team to do so," explained Adam Morawiec, VP of Business Development at Evatronix, in this SYS-CON.tv interview at @ThingsExpo, held Oct 31 – Nov 2, 2017, at the Santa Clara Convention Center in Santa Clara, CA.
To get the most out of their data, successful companies are not focusing on queries and data lakes, they are actively integrating analytics into their operations with a data-first application development approach. Real-time adjustments to improve revenues, reduce costs, or mitigate risk rely on applications that minimize latency on a variety of data sources. In his session at @BigDataExpo, Jack Norris, Senior Vice President, Data and Applications at MapR Technologies, reviewed best practices to ...
"ZeroStack is a startup in Silicon Valley. We're solving a very interesting problem around bringing public cloud convenience with private cloud control for enterprises and mid-size companies," explained Kamesh Pemmaraju, VP of Product Management at ZeroStack, in this SYS-CON.tv interview at 21st Cloud Expo, held Oct 31 – Nov 2, 2017, at the Santa Clara Convention Center in Santa Clara, CA.
Large industrial manufacturing organizations are adopting the agile principles of cloud software companies. The industrial manufacturing development process has not scaled over time. Now that design CAD teams are geographically distributed, centralizing their work is key. With large multi-gigabyte projects, outdated tools have stifled industrial team agility, time-to-market milestones, and impacted P&L stakeholders.
Enterprises are adopting Kubernetes to accelerate the development and the delivery of cloud-native applications. However, sharing a Kubernetes cluster between members of the same team can be challenging. And, sharing clusters across multiple teams is even harder. Kubernetes offers several constructs to help implement segmentation and isolation. However, these primitives can be complex to understand and apply. As a result, it’s becoming common for enterprises to end up with several clusters. Thi...
"Infoblox does DNS, DHCP and IP address management for not only enterprise networks but cloud networks as well. Customers are looking for a single platform that can extend not only in their private enterprise environment but private cloud, public cloud, tracking all the IP space and everything that is going on in that environment," explained Steve Salo, Principal Systems Engineer at Infoblox, in this SYS-CON.tv interview at 21st Cloud Expo, held Oct 31 – Nov 2, 2017, at the Santa Clara Conventio...
In his session at 21st Cloud Expo, James Henry, Co-CEO/CTO of Calgary Scientific Inc., introduced you to the challenges, solutions and benefits of training AI systems to solve visual problems with an emphasis on improving AIs with continuous training in the field. He explored applications in several industries and discussed technologies that allow the deployment of advanced visualization solutions to the cloud.
The question before companies today is not whether to become intelligent, it’s a question of how and how fast. The key is to adopt and deploy an intelligent application strategy while simultaneously preparing to scale that intelligence. In her session at 21st Cloud Expo, Sangeeta Chakraborty, Chief Customer Officer at Ayasdi, provided a tactical framework to become a truly intelligent enterprise, including how to identify the right applications for AI, how to build a Center of Excellence to oper...
"IBM is really all in on blockchain. We take a look at sort of the history of blockchain ledger technologies. It started out with bitcoin, Ethereum, and IBM evaluated these particular blockchain technologies and found they were anonymous and permissionless and that many companies were looking for permissioned blockchain," stated René Bostic, Technical VP of the IBM Cloud Unit in North America, in this SYS-CON.tv interview at 21st Cloud Expo, held Oct 31 – Nov 2, 2017, at the Santa Clara Conventi...
In his session at 21st Cloud Expo, Carl J. Levine, Senior Technical Evangelist for NS1, will objectively discuss how DNS is used to solve Digital Transformation challenges in large SaaS applications, CDNs, AdTech platforms, and other demanding use cases. Carl J. Levine is the Senior Technical Evangelist for NS1. A veteran of the Internet Infrastructure space, he has over a decade of experience with startups, networking protocols and Internet infrastructure, combined with the unique ability to it...
22nd International Cloud Expo, taking place June 5-7, 2018, at the Javits Center in New York City, NY, and co-located with the 1st DXWorld Expo will feature technical sessions from a rock star conference faculty and the leading industry players in the world. Cloud computing is now being embraced by a majority of enterprises of all sizes. Yesterday's debate about public vs. private has transformed into the reality of hybrid cloud: a recent survey shows that 74% of enterprises have a hybrid cloud ...
"Cloud Academy is an enterprise training platform for the cloud, specifically public clouds. We offer guided learning experiences on AWS, Azure, Google Cloud and all the surrounding methodologies and technologies that you need to know and your teams need to know in order to leverage the full benefits of the cloud," explained Alex Brower, VP of Marketing at Cloud Academy, in this SYS-CON.tv interview at 21st Cloud Expo, held Oct 31 – Nov 2, 2017, at the Santa Clara Convention Center in Santa Clar...
Gemini is Yahoo’s native and search advertising platform. To ensure the quality of a complex distributed system that spans multiple products and components and across various desktop websites and mobile app and web experiences – both Yahoo owned and operated and third-party syndication (supply), with complex interaction with more than a billion users and numerous advertisers globally (demand) – it becomes imperative to automate a set of end-to-end tests 24x7 to detect bugs and regression. In th...