System Design Guide for Dell PowerFlex family Flashcards
What is a benefit of two layer vs HCI build?
can help in lowering Enterprise Software licensing cost - only applied to compute nodes compared to HCI where it’s included in all the nodes
can make day 2 operations easier
allow external SDC nodes to connect to PowerFlex
How does data traffic work in an HCI build?
SDC to SDS (front end) and SDS to SDS (back end) data traffic runs on shared VLANs
How does deployment work in a hybrid build?
HCI nodes deployed separately from SO/CO nodes
recommendation is to have all homogenous nodes in the same PD to ensure consistent performance
How is storage consumed in a hybrid build?
SDCs in HC and CO nodes can consume storage from either or both PDs (HC or SO nodes)
How does ESXi deployment work on PowerFlex?
supported on HCI and CO nodes w/ full deployment and LCM through PFxM
How does non-ESXi deployment work on PowerFlex?
RHEL/CentOS/SUSE can be deployed on CO nodes through PFxM
customers responsible for providing OS Image (supported only deployment no LCM)
What is customer responsible for specifically to RHEL deployments on PowerFlex?
for configuring hypervisors such as KVM/THV on top of RHEL
What is included w/ PowerFlex Rack?
Smart PDUs
IPI
PFlex Management Controller
PFlex HW/SW
Networking
What OSs are supported specifically on PFlex Rack if using NVMe/TCP?
RHEL
ESXi
SLES
What is the minimum SO nodes for PFlex Rack?
4 SO nodes
How is networking built out in PFlex Rack?
each cabinet has redundant acc/agg or leaf/spine switches
pair are installed in first cabinet/configured in correct topology
if more than one cabinet exists the switches can be spread across or put in other cabinets
What vendors does PFlex Rack support?
Acc/Agg - Cisco and Dell
Leaf/Spine - Cisco
What is the support level on PFlex Rack?
guaranteed one call support
Release Certification Matrix upgrade of code upgrade to all components
What are the file node sizes on PowerFlex?
small - 2 x 12 cores
medium - 2 x 16 cores
large - 2 x 28 cores
19.2-76.8TB
128-256GB
What are the specs for NSX-T nodes?
2 x 14 cores
256GB
5 x 800GB SSD
6 x 25 GbE
PERC H755
What is the networking recommendation for all PFlex Rack builds?
select Leaf-Spine option (especially for high performance or greater scaling)
When can you use acc/agg networking topology in PFlex Rack builds?
when high performance/scalability aren’t big factors
when PDs are not crossing multiple pairs of ToRs
How are Access switches connected?
all nodes including management controller nodes connect to access switches
How many nodes can an access switch pair support?
24 nodes (1RU or 2RU)
physically 42U rack can only support 12 2U nodes
How are Aggregation Switches connected?
uplinks to customer’s core network (10/40/100Gbps) links
connect to acc switches using dual (100Gbps) uplinks
What is not supported by NSX-T nodes?
SDC and SDS traffic across it
What customer provided infrastructure is necessary for ESXi and Hyper-V deployments?
Windows Server 2016/2019
vCenter Server
What are the minimum requirements for customers that want to use their own switches for PowerFLex Appliance?
less than or equal to 1ms round trip time latency
min 4 x 25GbE SFP28 per node
min 2 x 100GbE QSFP28 uplinks per switch
Jumbo Frames
VLANs
802.1 VLAN Encapsulation
What is true about Partial Network Automation?
only available on PowerFlex Appliance
What is important to tell customers about partial network automation?
need to be aware of impact and need of peer-link bandwidth
switch peerlink bandwidth must be greater than or equal to uplink bandwidth (400G Peerlink bandwidth recommended)
What are the hardware components of the PFlex Management Controller?
controller nodes
management switches
software defined storage
system management software components
What are the controller nodes in the PFlex Management Controller?
based on PE Servers - use PFlex to provide a reliable and highly available storage cluster for the management plane
separate from the PFlex production nodes
What are the two configurations for controller nodes?
large
small
depends on compute/storage capacity
What are the management switches in PFlex Management Controller?
Cisco or Dell for OOB Management
What is the automation difference between Full and partial network automation?
PFxM automates server facing ports and node OS networking in FNA
PFxM automates node OS networking only in PNA
How does automation of server facing ports work in FNA?
once physical connection to OOB management/server facing ports - PFxM can assign IP address, bonding, port speed, port changes etc.
How does automation of OS networking work?
OS network = VMkernel, port groups, distributed switches, VLANs etc.
When would a customer be capable of deploying their own network for PowerFlex?
proficient in configuring networks
have access to network experts
must comply with specific security measures outside of PowerFlex’s best practices
How much space is reserved in PowerFlex Rack for nodes?
28RU
How much space is reserved in PowerFlex Rack for support shelves?
1RU support shelf reserved for every 8RU used for equipment
What is the sizing rule for CO nodes running ESXi?
2 node minimum per CO service in PFxM
What is the rule w/ Intel and AMD CO nodes in an ESXi environment?
can’t mix since vMotion will fail when crossing between Intel and AMD
What is the formula to calculate the number of nodes you can house in a PowerFlex rack cabinet?
1.125 * X + 2.25 * Y <= 28
Where
X = # of 1 RU nodes
Y = # of 2 RU nodes
What is the compatability between PFlex Appliance and other offerings?
must only use PFlex appliance nodes within appliance systems
Rack Nodes and ready nodes not compatible and not supported by appliance systems
What does it mean when PDs should be homogenous?
Node Type (SO/HC)
Disk type (SAS, SATA, NVMe)
Disk count (number of disks)
Disk size (GB)
What is the storage controller in SO and HC nodes?
HBA355
specifically for SSDs
What is the available power option on PowerFlex?
DC
What is a networking limitation of PowerFlex?
can’t mix network vendors in a single node
What is important to know about using NVMe drives?
dual CPU nodes required
What is important to know about using NVDIMMs?
Can be supported in single CPU nodes
Why is R840 not used for SO nodes?
quad CPU would be of no benefit since SO nodes service no compute
What OS must all SO nodes be running?
RHEL or Embedded PFlex OS
What software components do file nodes have?
FSN and SDC
How much RAM should you have for SO nodes over 51.5TBs?
32GB NVDIMM
56GB RAM
How much RAM should you have for SO nodes between 51.5 - 96TBs?
64GB NVDIMM
56-73GB RAM
How much RAM should you have for SO nodes between 96-128TBs?
96GB NVDIMM
90GB RAM
How much RAM does the MDM take?
5.4GB
How much memory does the OS take up in SO nodes?
1GB
What are the components of the Intelligent Physical Infrastructure (IPI)?
Smart PDUs
Physical Cabinets
Access Controls
Associated Cables
What are Smart PDUs?
work in parrallel w/ v4.5 cabinets
PDU connections to the IPI are pre-cabled w/ the v4.5 cabinet and power data is accessible through the IPI GUI
How can Smart PDUs be mounted?
whip up or whip down configurations
How are Smart PDUs/racks connected?
single Cat 5/6 connection and single IP address to the management switch
up to 3 pairs of PDUs
What are the dimensions on v4.5 racks?
27in wide
47 in deep
How many PDUs can go in a cabinet in PowerFlex Rack?
2-6
How many outlets per PDU?
36 - 24 10amp - 12 16amp
What are the hardware components of the PowerFlex Management Controller?
PFlex Rack Management Controller Nodes
NSX-T Edge Node
Management Switch
What are PFlex Rack Management Controller Nodes?
PowerEdge servers designed to provide compute power/storage capacity to run software components of PFlex Rack management
What are management switches connected to?
controller nodes
PFlex Nodes
NSX-T edge nodes
IPI v4.5 and PDUs
What is the management switch used by PowerFlex Rack in Cisco configurations?
Cisco 9238GC-X
1G
one recommended per rack
How many management switches should you have for 4-32 nodes?
1
How many management switches should you have for 33-78 nodes?
2
How many management switches should you have for 79-123 nodes?
3
How many management switches should you have for 123 and over nodes?
add 1 more switch every 45 nodes
Why can the first management switch only support up to 32 nodes?
since it’s partially consumed by management controller nodes, IPI appliance, PDUs, and customer uplink
What is the main benefit of having an OOB management network?
you can still access a switch even if a part of the network between you and the switch goes down
How are management switches connected?
one cable from the management switch to each device on the network to be managed
connected to OOB ports on switches
connected to iDRAC ports on servers
What is the management switch used by PowerFlex Rack in Dell configurations?
S4148T
What is the management controller requirement for PowerFlex appliance configurations?
single management node based on R650/640
multiple drives paired to internal RAID controller for redundancy
What is the general network setup in PowerFlex Rack for ToR Switches?
2 in each rack
can support up to 24 nodes and controllers
What are the standard connections for PowerFlex Rack nodes?
5
first 4 connected to ToR (access/leaf) via DAC (25G) or AOC (100G) cables
last connection to the management switch
What are the standard connections for PowerFlex Rack Management Controller?
6
first 4 connected to ToR (access/leaf) via DAC (25G) or AOC (100G) cables
last 2 connection to the management switch
What is the benefit of having dual networks in PowerFlex Rack?
enables Dell to install PFlex rack into environment w/ customer’s existing software defined network
What is the hardware requirement for dual network PowerFlex Rack?
two new acc switches
two additional ports per host
distributed virtual switch to funnel customer traffic into the SDN
What nodes are affected by the dual network design?
HC and CO nodes
SO nodes don’t require changes to function within dual network design
How many data networks should you design for dual network design?
2
can support up to 8 but more than 2 leads to design errors
What are the standard connections for PowerFlex HC and SO nodes in a dual network design?
7
first 4 connected to ToR (access/leaf) via DAC (25G) or AOC (100G) cables
next 2 connected to customer ACI leaves through DAC/AOC/25G optics
last connection to the management switch
How is L2 networking handled in an acc/agg configuration?
agg switches handle L2 connectivity down to all nodes
What protocols can agg switches run?
L2 or L3 protocols
L2 is default
What is the recommendation when running an acc/agg network topology?
multi-chassis ling aggregation (MLAG) recommended to ensure high bandwidth for expected performance and resiliency
customer can still choose no MLAG if they want
How is MLAG referred to by different vendors?
VLT - Dell
VPC - Cisco
What is the PD recommendation when deploying acc/agg topology?
PDs should reside under a single pair of acc switches due to amount of east/west traffic
for greater east/west traffic leaf spine option is recommended if PDs need to cross ToR boundaries
What types of spine/leaf configurations are supported on PowerFlex?
3-spine
6-spine
What effect does adding more spines to spine leaf topology do?
does not enable a larger environment but will increase amount of east west traffic that environment can support
What are the types of switches in spine leaf topology?
spine
leaf (ToR)
border leaf (BL)
What are spine switches?
backbone of network responsible for interconnecting leaves and borders
What are leaf switches?
provide connectivity to nodes/controllers in environment
What are border leaf switches?
provides connectivity to the customer network
What leaf spine topology is supported by Dell?
Cisco
SONiC Leaf w/ Dell Switches (PNA only)
What is the standard of a 25G leaf spine configuration?
4 25G connections to each node w/ 100G uplink to spines
supports up to 24 nodes per leaf pair
What is the standard of a 100G leaf spine configuration?
4 100G connections to each node
100G to 10G connections to controllers
supports up to 8-16 nodes per leaf pair depending the switch model
What PowerFlex offerings support NSX-T?
Rack and Appliance
What additional components are needed to run NSX-T in PowerFlex environment?
additional 4th controller to support NSX-T Manager software
2 or more NSX Edge nodes to run NSX-T Edge Service Gateway instances
What are the configuration options for controller nodes on PowerFlex?
Large or Small
Always recommended to do large
What is the instance limit of NSX-T Edge Service Gateway?
each Edge Node can run max of two ESG VMs
How are the NSX-T Managment and Control Plane mounted?
host on NSX Manager VMs that live on Management Controller cluster
How do NSX-T Edge nodes work?
in band nodes that have app/workload traffic forwarded under the control of the NSX-T controllers
What is the data path of NSX-T Edge nodes?
data runs through ESG
all outbound/inter-VLAN traffic will pass through ESGs in route to external/ between zones in the NSX-T environment
why they have dedicated ports
How is the customer’s physical network connected to the ESGs in an NSX-T setup?
external network not physically attached
connected via uplinks to from the Agg or Spine Layer
What is a limitation with an appliance configuration using NSX-T Edge Nodes?
appliance do not allow customers to bring their own management controllers
What does the multinetwork feature of PowerFlex allow?
customer can have different VLANs and subnets in each cabinet and PowerFlex can span them
multiple VLANs and subnets can be assigned per role
What is the node minimum for PFlex Rack?
4 HC nodes
recommendation of 6
How many nodes can PFlex Rack scale to?
168 in acc/agg
1000 in spine/leaf
What is the PD/SP limit for PFlex Rack?
32 nodes per PD
300 drives per SP
What is best practice for PD and SP configuration in PFlex Rack?
1:2 ratio for PD:SP
What are the networking minimums for Pflex Rack?
1 pair of agg or 3 spines
1 pair of acc or leafs
How many management controller nodes do you need in PFlex Rack?
3 minimum
1 management switch minimum
What is the minimum for PFlex nodes for file?
2-16
What is the shared storage pool available for PowerFlex file to work?
5.5TBs
What is the max nodes that can fit in a single cabinet in PowerFlex Rack?
21 1RU nodes
Why does the first PFlex Rack cabinet have agg switches built in if its an agg acc configuration?
to help smooth expansion beyond one cabinet (a pair of acc switches)
What is the max nodes that can fit in a 2 rack PowerFlex Rack configuration?
45 1RU nodes
What is the minimum node configuration for PFlex appliance?
4 SO or HC nodes
6 recommended (min of 6 for a PD)
What are the minimum network configurations for PowerFlex appliance?
1 pair of acc/leaf switches
min 1 management switch (up to 47 nodes)
What kind of replication is available on PowerFlex?
asynchronous storage based replication
What is the default MDM cluster mode for a 4 node minimum configuration?
3 MDMs (primary, secondary, tiebreaker)
two copies of repository so can withstand one MDM cluster failure
What is the default MDM cluster mode for a configuration 5 nodes or more?
5 MDMs (primary, 2 secondary, 2 tiebreakers)
three copies of repository so can withstand two MDM cluster failures
How do the MDM, SDS, and SDC interact in a VMWare environment?
MDM and SDS are hosted inside a dedicated VM (SVM)
SDC installed in ESXi
What is PowerFlex capacity licensing based on?
Raw capacity - price per TB get’s cheaper the higher capacity tier the licensing is
What is the static route setup for internal SDCs in PowerFlex Rack?
layer 3 routing from SDC to SDS enables IP subnet separation between groups of SDC’s to a common SDS pool
enables multi-tenancy capability on PFlex Rack
What kind of multi-site replication is available on PowerFlex?
up to 5 separate sites on 4.0 and up releases
multi-site multi-directional replication also supported
What happens if an SDR fails?
there is a minimum number of SDRs deployed on a system
therefore MDM directs SDC to send the IO to available SDR
What are the rules for SDRs?
SDRs are specific to a PD and each PD will require a minimum number of SDRs for availability
SDRs must be installed along with SDSs
What kind of checksum is supported MG pools?
inflight checksum
does not support persistent checksum
What are MG pools best suited for?
workloads w/ high performance and low latency requirements
does not support data compression
supports thin or thick LUNs
How can pools be layed out in an SDS?
FG and MG pools can live side by side in each SDS
volumes can be migrated between the two layouts
What hardware is required to deploy FG pools?
can only be used w/ SSD/NVMe
require an acceleration pool with NVDIMMs/SPDM
What kind of compression is supported on FG pools?
in line compression
What kind of workloads are supported on FG pools?
only thin LUNs
best choice for compressable data and space efficiency
recommended when thin provisioning and snapshots are key requirements
What kind of checksum is performed on FG pools?
persistent and in flight checksum
What is the design rule for SP layout?
PDs can have multiple pools of MG but multiple FG Pools or FG pools with MG pools in same PD is not recommended
What is the rule for sizing fault sets?
minimum of 2 nodes per FS but recommended of 4
What is the rule for replication consistency groups?
one RCG consists of one source and up to 4 targets
source and target must reside in separate SPs
source and target SP should be of identical size
How much space should be reserved on SPs for replication journals?
10%
What is the performance rule for asynchronous replication in regard to SPs?
performance character of any SP where a journal volume resides must meet or exceed the performance character of the SP where the replicated volume resides
MG/MG or MG/FG or FG/FG allowed basically
What are the RPO min and max on PowerFlex?
15sec - 1hr on 4.0 or greater update
What is the bandwidth rule for replicated volumes on PowerFlex?
replicated volumes require 3 times the network bandwidth vs non-replicated volumes
sustained bandwidth of all volumes being replicated should not exceed 80% of available WAN bandwidth
What is the latency limit for asynchronous remote replication?
200ms
What are the different configuration options for file nodes?
small - 24 cores - 128GB RAM - 4 x 25G
medium - 32 cores - 256GB RAM - 4 x 25G
large - 32 cores - 1024GB RAM - 4 x 25/100G
What do NAS Servers allow for PowerFlex to enable?
Multitenancy
file systems assigned to NAS Servers upon creation and can’t be moved to another NAS Server
How are new NAS Servers assigned in PowerFlex?
round robin fashion
How are NAS Servers provided high availability on PowerFlex?
primary node acts as marker to indicate to node that the NAS Server should be running on
once provisioned the primary node for NAS Server never changes
backup node in place NAS Server is backed up for fault tolerance purposes (Created automatically through node balancing algorithm)
How is Powerflex file managed?
PowerFlex Unified Manager GUI
What is the PowerFlex Gateway?
serves as the PowerFlex REST API gateway and SNMP trap sender
provisioned as separate VM on PFMC or customer provided hw (appliance config only)
What is true about deploying the PFlex Gateway in modern configurations?
deploying Gateway only for v3.8 or below
not required for v4.0 and above - two gateways installed automatically
What is the PowerFlex Presentation Server?
provisioned as a separate VM which provides the UI
only required to be installed in v3.8 or below - automatically in place for 4.0 and above installations
How is PFxM deployed in v3.8 or earlier?
supplied as a virtual appliance in OVA format
runs as single VM and must be deployed on ESXi
single instance of PFxM 3.8 can be used to manage multiple clusters
still deploy separate instances of PFlex Gateway to presentation server
How is PFxM deployed in v4.0 or later?
PFxM 4.0 is now container based architecture deployed across 3 management VMs
supplied VM based installer must be used to deploy the cluster
integrates functionality of both Gateway and presentation server
single instance of PFxM can be used to manage one cluster
What features are only available with PFxM 4.0 and above?
NVMe/TCP
file services
How is PFxM deployed in Rack configuration?
deployed on the 3 node management controllers
How is PFxM deployed in Appliance configurations?
can be on Dell provided single management controller/multi-node controller or customer provided hardware
What are the resource allocations for PFxM 4.0 or above if customer is using own hardware?
42 cores
96GB
1.8TB
When can a customer deploy PFxM 4.0 on their own management infrastructure?
ESXi or KVM environments
How would customers integrate their management infrastructure if they’re already ESXi or KVM customers?
customer deploys our ESLES VMs on their hypervisor to run management
images provided as OVA for ESXi
images provided as QCOW for KVM
What is vCenter used for in ESXi configurations?
to manage the management control plane cluster VMs and nodes as well as customer VM environment/nodes
deployed in VHCA (vCenter High Availability) mode
What is Dell Secure Connect Gateway?
enterprise monitoring application that is delivered as an appliance or software
monitors devices and proactively detects hw issues
What is a jump host?
VM that resides on the management controller
provides working environment with access to all PFlex Rack internal networks and troubleshooting ability
What is important to know about customers that want to use NSX-T?
we don’t sell or install NSX-T as part of PowerFlex system (Customer acquires licenses from VMWare)
What are the configuration min and max for NSX-T edge nodes?
2-8
What are the storage configuration options for NSX-T edge nodes?
RAID 10/vSAN
What is the role of CloudLink?
integrates w/ PFlex to orchestrate D@RE and key management
How does CloudLink work?
selectively enables device level encryption using software based encryption
How is CloudLink deployed on Rack configurations?
CloudLink Center deployed on PFMC
How is CloudLink deployed on Appliance configurations?
customer provides management platform and CloudLink Agent deployed by PFxM on HC and SO nodes
How is CloudLink licensed?
by raw TB
What are the SecureVM CloudLink licenses?
enable use of CloudLink to encrypt VM or bare metal volumes
What are KMIP KMS licenses for CloudLink?
enable use of CloudLink as the key store if customer wants to use native vSphere encryption and need secure key location
How is PFMC deployed in a Rack configuration?
3 node offering w/ HBA355
How is PFMC deployed in a Appliance configuration?
single node (HBA755) or 3 node
What are the rules w/ acceleration pools?
acceleration pool can serve multiple SPs but a single SP can’t use multiple acceleration pools
an acceleration device can serve multiple SDS devices but SDS devices can’t span across multiple acceleration devices
What can you use as an alternate to FSN nodes?
PowerScale nodes in a Rack configuration
no orchestration or automation through PFxM if using PScale nodes for file data
What is true about PowerFlex scalability with legacy nodes?
14, 15, 16 gen nodes can reside in the same cluster
What is the maximum node limit for a PD in PFlex Rack?
30 nodes
What are the node limits per rack on PFlex Rack?
21 1U nodes - (first rack
24 - 1U
12 - 2U
What is the management switch requirement for smaller PFlex Rack clusters?
recommended to have one per cabinet but not required if node count is lower than 24
What are the max switches supported by PFlex Rack?
only one pair of agg switches under acc/agg topology
3/6 spines
pair of acc/leaf switches required per cabinet
What are the max node counts for each spine configuration for PFlex Rack?
3 spine - 144 leafs - 168 nodes (1RU)
6 spine - 14 leafs - 168 nodes (1RU)
How can you achieve more than 168 node max in PFlex Rack?
using switches w/ higher port density
What is the recommendation for spine switch installation for PFlex Rack?
install spine switches in separate cabinets to ensure resiliency
When should you use IMM instead of PMM?
if you plan to place a node in maintenance mode for less than 30 minutes
What mode is the cluster put into when IMM is initialized?
degraded mode - should not leave node in IMM once completed
What is the capacity calculation for journal capacity requirements?
must equal at least 28GB of the SP per SDR
at least 5% of the usable capacity of the replicated volumes
of SDRs x 28
What is the rule when sizing journal capacity by RPO?
as RPO decreases amount of journal capacity needs to increase
What is the rule when sizing journal capacity by RTO?
as RTO increases amount of journal capacity needs to increase
What is a replication consistency group?
logical container for volumes whose application data needs to be replicated consistently to each other
What is defined by the RCG?
RPO
Consistency Mode
Pause Mode
Freeze Mode
What type of nodes support asynchronous replication?
SO nodes and ESXi based HCI nodes