Cisco HyperFlex Design, Performance and Benchmarking for All-Flash and NVMe Clusters
Silvo Lipovše, sistemski inženjer
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 3
Agenda
BRK
What’s New
All-Flash System Design
Benchmarking Tools
Performance Testing
Conclusion
What’s New in HyperFlex
4
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 5
What’s New With Cisco HyperFlex 2.6?
All-Flash Now Supports NVMe Caching Disks Support for Self Encrypting Disks (SEDs)
Cisco Intersight Cloud-Based Management
HyperFlex Connect Native HTML5 GUI with Role Based Access Control
Native Snapshot Based Replication Expanded Compute-Only Node Support
M5 Generation Server Models
Mixed Clusters of M4 and M5 Servers, Higher Capacities and
Larger Scale
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 6
New Disk OptionsNon-Volatile Memory Express (NVMe)
Replaces SAS write caching disk
More efficient streamlined protocol
Uses PCI-Express instead of legacy serial bus
Multi-queue, more commands per queue, high parallelism
Higher performance, lower latency
Self Encrypting Disk (SED)
Used for write caching disk and capacity disks
Disk hardware encrypts all data written to the device
Can use UCSM derived local encryption keys, or keys from a centralized KMS via KMIP
Secure erase, re-key
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 7
Cisco HyperFlex All Flash Configurations HXAF240c Cluster
Capacity-Optimized 3–16 Node Cluster
Per-NodeCaching Device: 1 x 800 GB SSD, 400 GB
NVMe, or 800 GB SEDCapacity Disks: up to 23 x 960 GB, 3.8 TB
SSD, or 800 GB SED Boot/Housekeeping: M.2 SSD + 120GB or
240GB SSD
HXAF220c Cluster
Smallest Footprint 3–16 Node Cluster
Per-NodeCaching Device: 1 x 800 GB SSD, 400 GB
NVMe, or 800 GB SEDCapacity Disks: 6 x 960 GB, 3.8 TB SSD,
or 800 GB SED Boot/Housekeeping: M.2 SSD + 120GB or
240GB SSD
Expanded Clusters
Compute-heavy Hybrid(Compute Bound Apps/VDI)
Up to 16 Compute NodesBlade or Rack
B200 M3/M4, B260 M4, B420 M4, B460 M4, C220 M3/M4, C240 M3/M4, and C460 M4
Local Disk, SD Card or SAN Boot
3-16 Node HXAF220 or HXAF240 Cluster
Standard UCS 2 & 4 Socket Servers
+
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 8
HyperFlex Native Data Protection and SecurityProtect against Site
Failure
Data-At-Rest Encryption thru Self-Encrypting Drives
Enterprise Key Management Support
Data-At-Rest Loss Prevention
Native Replication of Snapshots across Sites
Disaster Recovery Protection
Regulatory Compliance
Certifications starting with FIPS, Common Criteria
Compliance for Data Privacy (GDPR, HIPAA, PCI-DSS etc.)
Secure Platform
Vulnerability Assessments and Hardening
Role-based access control and Auditing
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 9
Cisco Intersight Cloud-Based Management
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 10
Native HyperFlex Connect HTML5 GUI
All-Flash System Design
11
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 12
Building on the Right FoundationCisco HX Data Platform
Built From the Ground Up for
Hyperconvergence
Distributed Log-Structured File
System Designed for Scale-out,
Distributed Storage
Advanced Data Services (Snapshots, Clones, Replication)
and Data Optimization (Inline Dedupe,
Compression) Without Trade-offs
Better Flash Endurance and
Disk Performance
Computing, Storage,
Networking, and Hypervisor Integration
No Reliance on Legacy
Filesystems or
Technology
Distributed File system
Local Disks
UniqueArchitecture
Local Disks Local Disks Local Disks
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 13
Software Modules Inside a Server
Controller VM Has Direct Control of Drives
VAAI Plugin Offloads Snapshots
and Clone OperationsIO Visor Module Presents
NFS to ESX and Stripes IO
DATASTORE/VOLUME
HYPERVISOR
HDD
HDD
SDD
SDD
SSD SSD SSD SSD
CONTROLLERVMVAAI
stHyp
VM VM VM VM VM
IO Visor
stHypervisorSvc Coordinates Native
Recovery Snapshots and Replication
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Read Cache
Log FS
L1
Read
Distributed Objects
All Flash
DRAM
Read Cache
14
Optimizing for All Flash: Read Path
Log FS
L2
L1
Read
Distributed Objects
Hybrid L1 (DRAM) Cache:• Metadata
and data • Helps with
small bursts
L2 (SSD) Cache:• De-duplicated
Read Cache• Use aggregate
performance of all Read Cache SSDs
SSD
DRAM
No L2 (SSD) Cache:• No flash on flash
advantage• Use aggregate
performance of all SSDs
SCVM RAM:• Unified capacity disk
index cache• Improvements by
avoiding additional index reads
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Technical Improvements in HX 2.5+ Updated Linux kernel in the controller VMs
– Gains of 20-30% in read performance
Optimized write cache destaging process– More consistent performance with less variability
Metadata Optimization– Less space consumption, fewer updates required, faster index lookups
Cleaner Improvements– Lower performance impact, less variability
Write Log Enhancements– In-line compression with no latency penalty, makes effective write log size larger
15
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
DATASTORE DATASTORE
16
Data Distribution and Non-Disruptive Operations
Stripes blocks of virtual disk files across servers using a hashing algorithm
Replicate one or two additional copies to other servers
Handle entire server or disk failures
Restore back to original number of copies Rebalance VMs and data post replacement Rolling “one-click” software upgrades
CONTROLLERHYPERVISOR
VM VM VM
CONTROLLERHYPERVISOR
VM VM VM
CONTROLLERHYPERVISOR
VM VM VM
CONTROLLERHYPERVISOR
VM VM VM
File.vmdk
D1 E1A1 B1 C1B2 A2 A3C2 C3 D2D3 E2E3 D1E1 B3 B3
EDCBA
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 17
Life of a Write IO: Write LogCONTROLLERVM
IOvisor
VMVM
ESXiDS
CONTROLLERVM
IOvisor
VMVM
ESXiDS
CONTROLLERVM
IOvisor
VMVM
ESXiDS
AA’’ A’
ACKA
1. Application writes data (A) to OS2. VM writes to VMDK on datastore3. Write intercepted by IOvisor and sent to primary controller for that VMDK segment and offset4. Primary data is compressed and committed to WL SSD5. Simultaneously, data replicated to two other nodes (RF3) and committed to their WL SSD6. Acknowledgement sent to VM7. Data is successfully committed and protected
B B’ B’’CC’’C’
D D’D’’
Write Log Capacity Used
Size: 8GB or 32GB Hybrid / 60GB AF
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 18
Life of a Write IO: Destage
AA’’ A’
1. When the active write log of a node is full, the active log becomes the secondary log and destages2. Primary data copies are deduplicated3. Data is then three way written (RF3) to SSD / HDD on 3 separate nodes4. After all data is successfully committed, data is purged from the secondary log on the caching SSD5. In hybrid systems, recent writes are added to the read cache area of the SSD during destage
B B’ B’’CC’’C’
D D’D’’
Write Log Capacity Used
CONTROLLERVM
IOvisor
VMVM
ESXiDS
CONTROLLERVM
IOvisor
VMVM
ESXiDS
CONTROLLERVM
IOvisor
VMVM
ESXiDS
A’A’’AA
BBB’B’’
CC’’C’ D’’ D D’
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
CONTROLLERVM
IOvisor
VMVM
ESXiDS
CONTROLLERVM
IOvisor
VMVM
ESXiDS
CONTROLLERVM
IOvisor
VMVM
ESXiDS
19
Life of a Read IO
ESXi
Read A
1. Application reads data from OS filesystem2. VM reads from VMDK on datastore3. Read intercepted by IOvisor and sent to primary controller for that VMDK segment and offset
a) First check active write log b) Next check passive write log (the duplicated write log of the other nodes)c) Look in L1 (DRAM) cached) Look in L2 (SSD) cache [Not in AF models]e) Retrieve from SSD / HDD – copied to L2 [Not in AF models] and L1, decompress & returned to IOvisor
4. Return data to VM
Active WL
Passive WL
L1
L2
SSD (AF) / HDD (Hybrid)
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 20
Dynamic Data Distribution• HX Data Platform stripes data across all nodes simultaneously,
leveraging cache across all SSDs for fast writes, and HDDs for capacity, via the low-latency UCS internal network
• Balanced space utilization, no data migration required following a VM migration, failure tolerance, and no hotspots
Systems Built on Conventional File Systems Write Locally, Then Replicate, Creating Performance Hotspots
CONTROLLERHYPERVISORHYPERVISOR CONTROLLERHYPERVISOR CONTROLLERHYPERVISOR
VM VMVM VM VMVM VM VMVM
HX Data Platform
VM VMVM
CONTROLLERCONTROLLER
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Scale Compute
HX Data Platform
Add NodesScale Capacity Within Nodes
HX Data Platform
CONTROLLERHYPERVISORCONTROLLERHYPERVISOR CONTROLLERHYPERVISOR CONTROLLERHYPERVISOR
IOVisorIOVisor
IOVisorIOVisor
IOVisorIOVisor
IOVisorIOVisor
VM VM VM VM
Non-HyperFlex Hosts Can Connect to
Storage with IOVisor
VM VMVM VM VMVM VM VMVM VM VMVM
IOVisor
VM VM
IOVisor
VM VM
21
Independent Scaling of Compute and Capacity
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 22
HyperFlex Capacity Considerations SSD reliability is significantly better than HDD; from 2-10 times lower annual
replacement rate in studies by major online providers.
HyperFlex installer no longer defaults to RF3 for an all-flash system, forcing the end user to make a decision.
Consider RF2 for all-flash systems that do not require triple data copies or multiple component redundancy.
RF2 shows 15-40% reduced write latencies in lab testing.
RF2 can provide clusters up to 591 TiB maximum capacity.
Keep overall cluster capacity consumed below 70% for best performance.
Benchmarking Tools
23
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Iometer Intel project now developed as an open source project
No active development in 3+ years
Primarily a Windows GUI tool, but ports exist for Linux and Mac OS X
Easy configuration of an access profile
Not a primary tool for our team any longer
Issues:– Not very easy to set up server/client multi node tests– Difficult to script runs on Windows– No configurable dedupable or compressible data patterns– Default patterns are highly dedupable which can give false results
24
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
FIO Written by Jens Axobe, the current Linux kernel block layer maintainer
Compiled binary install for most Linux distributions
Configurations to test are written in text based job files
Highly customizable
FIO can be configured to run at startup, then a master machine can start tests remotely
Issues:– FIO must be installed on all testing machines– If FIO is not already running, you cannot start a multi-node test– Output can quickly get cluttered and hard to read
25
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Vdbench Written by Henk Vandenbergh at Oracle
Java application that can run on almost all platforms with a JRE
Configurations to test are written in text based job files
Highly customizable
Easy server/client job execution via SSH or RSH
Easy to read output in HTML files
Has become the standard tool for the Cisco HX BU TME team
Issues:– Thread and JVM count can make calculating the OIO per disk difficult to calculate
26
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 27
Vdbench Output
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Cisco HX Bench Cisco developed tool to automate many facets of benchmark testing
Download at https://hyperflexsizer.cloudapps.cisco.com (CCO login required)
Deploys a master test controller VM from an OVA file on your HX cluster
HTML GUI makes deployment of VMs and test configuration easy
Uses vdbench for the testing
Use built in test scenarios or build your own
Automatically deploys lightweight Ubuntu VMs as the test load generators
Test results and charts are customizable and even easier to view than the default HTML output from vdbench
28
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 29
Cisco HX Bench User Interface
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
Others HCIBench
– VMware automation wrapper that executes vdbench on Photon VMs– Results are identical to vdbench, somewhat easier to use but not as flexible
HammerDB– Used internally for database workload qualifications
SQLIO, SLOB, Jetstress, etc.– Application specific tests not used for general performance characterization
Mixed Workload Tools– Examples: Weathervane, VMmark
30
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 33
Vdbench Example Job FilesVdbench
dedupunit=8kdedupratio=1.25compratio=1.4hd=default,user=root,shell=ssh,jvms=1hd=vdbench1,system=192.168.100.11hd=vdbench2,system=192.168.100.12hd=vdbench3,system=192.168.100.13hd=vdbench4,system=192.168.100.14
sd=default,openflags=o_direct,threads=16sd=sd1,host=vdbench*,lun=/dev/sdb
wd=wd1,sd=sd*,xfersize=8k,rdpct=70,seekpct=100
rd=rd1,wd=wd1,iorate=max,elapsed=5m,interval=5
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 34
Vdbench Example Job ExecutionEnsure each load generator VM has the vdbench java executable in the same location, and copy SSH keys so SSH logins happen without a password.
Since the vdbench job file specifies the clients as well, just run the test:
vdbench –f 7030_4k.vdb –o /var/www/html/
• Easy to loop multiple runs in sequence:for i in {1..3} ; do vdbench –f 7030_4k.vdb –o /var/www/html/ && sleep 3600 ; done
• Simple scheduling of runs with Linux at:# at 1am tomorrow
at> vdbench -f 7030_4k.vdb -o /var/www/html/output/7030_4k.tod
at> <EOT> (Press CTRL+d to end the job creation)
Performance Test Results
35
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 36
HX Hybrid to All-Flash Performance Comparison
0,0
1,0
2,0
3,0
4,0
5,0
6,0
7,0
0.4GB 100GB 800GB 2TB
Perf
orm
ance
Mu
ltip
le
Effective Working Set Size per NodeTests done on 4 x HX240
Performance With Varying Working Set
All Flash Performance Multiple
HX240 hybrid read cache 1.2 TB per node
• AF offers performance improvements and lower latency even at small working sets
• AF offers huge performance and latency improvements for working sets that do not fit in the hybrid model read cache
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 37
Base Configurations for Vdbench Tests
HX All-Flash HX All-Flash + 40GbE HX All-Flash + 40GbE and NVMe HX All-Flash M5
Version HX 2.0.1a HX 2.0.1a HX 2.5.1a HX 2.6.1b
HW Configuration
8 x HXAF240 M4, each with:• WL: 800G SSD• Data: 10 x 960G• VIC 1227• E5-2690v4• 512 GB RAM
8 x HXAF240 M4, each with:• WL: 800G SSD• Data: 10 x 960G• VIC 1387• E5-2690v4• 512 GB RAM
8 x HXAF240 M4, each with:• WL: 400G NVMe• Data: 10 x 960G• VIC 1387• E5-2690v4• 512 GB RAM
8 x HXAF240 M5, each with:• WL: 400G NVMe• Data: 10 x 960G• VIC 1387• Xeon 6154• 384 GB RAM
Benchmark Setup
• Vdbench 5.04.06• 8 Linux VMs per host (64 total)• Each VM has 1 x 200GB virtual disk, HX filesystem ~60% full (12.4 TB working set)• Priming: 64K writes 100% sequential• Tests run 3 times and averaged
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 38
Benchmark 1 Performance Results Using Vdbench Tests8K 100% Sequential Write Workload
• Target Write Latency < 10ms• 16 OIO Per Disk• 27% Latency Reduction and
IOPs Gain with 40 GbE• 6% Latency Reduction and
IOPs Gain on 2.5 + NVMe• 8% Latency Reduction and
IOPs Gain with HX 2.6 on M596.426
122.514130.015
140.397
10,618,35 7,87 7,30
0,002,004,006,008,0010,0012,0014,0016,0018,0020,00
0
20.000
40.000
60.000
80.000
100.000
120.000
140.000
160.000
2.0 2.0 + 40GbE 2.5 + 40GbE 2.6 M5
LatencyIOPs
Maximum Sequential WriteIOPs Latency
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 39
Benchmark 2 Performance Results Using Vdbench Tests8K 100% Random Read Workload
• Target Read Latency < 5ms• 32 OIO Per Disk• 7% Latency Reduction and
IOPs Gain with 40 GbE• 23% Latency Reduction and
IOPs Gain on 2.5 + NVMe• 3% Latency Reduction and
IOPs Gain with HX 2.6 on M5
389.526415.449
511.992 529.241
5,26 4,934,00 3,87
0,002,004,006,008,0010,0012,0014,0016,0018,0020,00
0
100.000
200.000
300.000
400.000
500.000
600.000
2.0 2.0 + 40GbE 2.5 + 40GbE 2.6 M5
LatencyIOPs
Maximum Random ReadIOPs Latency
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 40
Benchmark 3Performance Results Using Vdbench Tests100% Random 70/30% R/W with Varying Block Sizes
4K 8K 16K 64K 256K
IOPs R Latency W Latency IOPs R Latency W Latency IOPs R Latency W Latency IOPs R Latency W Latency IOPs R Latency W Latency
2.0 150,673 2.12 6.36 130,511 2.40 7.47 92,130 3.49 10.35 34,965 10.75 23.70 10,079 45.60 62.80
2.0 + 40 GbE 152,434 2.22 6.00 142,721 2.36 6.47 111,905 3.13 8.03 51,146 6.96 17.16 19,712 11.68 59.28
2.5 + 40 GbE 161,131 2.24 5.37 160,290 2.09 5.75 136,433 2.45 6.78 56,339 6.30 15.74 20,036 9.17 63.61
2.6 M5 180,037 1.72 5.49 173,193 1.91 5.39 140,842 2.36 6.61 61,227 5.45 15.18 20,732 8.96 61.37
• Target Read Latency < 5ms and Write Latency < 10ms for 4-16K blocks• 8 OIO Per Disk• ~6% Latency Reduction and IOPs Gain with 40 GbE at 4-8K, 50-100% Gains for Larger Blocks• 10-22% Latency Reductions and IOPs Gains with HyperFlex 2.5 and NVMe Caching Disks• 3-11% Latency Reductions and IOPs Gains with HyperFlex 2.6 on M5 Generation Servers
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 42
Benchmark 4Performance Results Using Vdbench Tests8K 100% Random 70/30% R/W IOPs Curve
• Curve test measures IOPs stability by calculating the standard deviation of IOPs as a percentage of the final average result, at various performance levels, starting with a 100% unthrottled test, then again at a percentage of that maximum
• Maximum speed, unthrottled tests will always have more variability due to more aggressive background tasks• 20% through 60% values are typically very low because the system is not being pushed very hard• 65% improvement in IOPs stability with 40 GbE at 80% of maximum HX 2.5 shows a further 78% improvement• HX 2.5 also shows a 54% improvement of IOPs stability in the 100% unthrottled test• HX 2.6 on M5 generation servers shows improvements in all tests, including 28% at 100%, and 45% at 80%
100% 20% 40% 60% 80%
IOPs IOPs STDV
Avg Latency IOPs IOPs
STDVAvg
Latency IOPs IOPs STDV
AvgLatency IOPs IOPs
STDVAvg
Latency IOPs IOPs STDV
AvgLatency
2.0 138,190 19.0% 3.71 27,666 0.5% 1.07 55,299 0.5% 1.57 82,960 0.5% 2.20 110,594 6.6% 3.14
2.0 + 40 GbE 151,568 18.6% 3.38 30,331 0.5% 0.99 60,621 0.5% 1.37 90,957 0.5% 2.08 121,250 2.3% 2.63
2.5 + 40 GbE 169,983 8.6% 3.01 34,031 0.4% 0.86 68,058 0.5% 1.17 102,023 0.5% 1.66 136,018 0.5% 2.28
2.6 M5 173,193 6.2% 2.95 34,699 0.2% 0.80 69,331 0.2% 1.21 103,953 0.2% 1.68 138,580 0.3% 2.38
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 43
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 45
Benchmark 5Performance Results Using Vdbench TestsI/O Blender 100% Random 70/30% R/W IOPs Curve
100% 20% 40% 60% 80%
IOPs IOPs STDV
Avg Latency IOPs IOPs
STDVAvg
Latency IOPs IOPs STDV
AvgLatency IOPs IOPs
STDVAvg
Latency IOPs IOPs STDV
AvgLatency
2.0 104,624 25.6% 4.89 20,967 0.5% 1.31 41,898 0.5% 1.72 62,825 0.5% 2.53 83,756 4.4% 3.89
2.0 + 40 GbE 114,354 22.2% 4.47 22,902 0.5% 1.04 45,792 0.5% 1.47 68,664 0.5% 2.47 91,435 3.1% 2.97
2.5 + 40 GbE 137,078 13.6% 3.73 27,466 0.5% 0.96 54,897 0.5% 1.45 82,292 0.5% 2.13 109,727 0.5% 2.77
2.6 M5 137,416 12.1% 3.72 27,496 0.2% 0.80 54,990 0.3% 1.49 82,456 0.3% 1.96 109,955 0.4% 2.89
• Test measures IOPs stability by calculating the standard deviation of IOPs as a percentage of the final average result, at various performance levels, starting with a 100% unthrottled test, then again at a percentage of that maximum
• Block Size Mix: 4K 40%, 8K 40%,16K 10%, 64K 10%• 20% through 60% values are typically very low because the system is not being pushed very hard• 30% improvement in IOPs stability with 40 GbE at 80% of maximum, HX 2.5 shows a further 84% improvement• HX 2.5 also shows a 39% improvement of IOPs stability in the 100% unthrottled test• HX 2.6 on M5 generation servers shows improvements in all tests, including 11% at 100%, and 20% at 80%
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia
0
5000
10000
15000
20000
25000
30000
1 7
13
19
25
31
37
43
49
55
61
67
73
79
85
91
97
10
3
10
9
11
5
12
1
12
7
13
3
13
9
14
5
15
1
15
7
16
3
16
9
17
5
18
1
Intervals
Vendor Y
VM1 VM2 VM3 VM4 VM5 VM6
VM7 VM8 VM9 VM10 VM11 VM12
48
Per VM Test Results VariationIOPS in a 70/30 Workload – 1 Hour Duration
0
5000
10000
15000
20000
25000
30000
1 8
15
22
29
36
43
50
57
64
71
78
85
92
99
106
113
120
127
134
141
148
155
162
169
176
Intervals
HX
VM1 VM2 VM3 VM4 VM5 VM6
VM7 VM8 VM9 VM10 VM11 VM12
19 - 21 March 2018 | Cisco Connect | Rovinj, Croatia 49
HyperFlex Performance Summary
HyperFlex shows large gains in performance by transitioning to 40 GbE and 3rd Gen FIs and VICsHyperFlex 2.5 + 40 GbE offers significant gains in IOPs, latency and stability for all test workloadsHyperFlex 2.6 on M5 shows further improvements in IOPs, latency and more consistent performanceHX All Flash, unlike competition, shows highly consistent performance across all VMs
BRK50