Server Hardware Flashcards
form factors
- variety of sizes and configurations of physical servers/components
- increasing data center capacity is possible using smaller server form factors
- must keep inventory to be aware of the form factors in the system environment
- virtual servers don’t have form factors
tower servers
- all server components are housed within a single case that can sit directly on the floor or on a desk
- standalone computer
tower server advantages
- can be easily moved without removing screws or sliding out of rack
- components are easy to find
- adding physical components is easy because of the amount of space inside the case
tower server disadvantages
- don’t scale well
- take up a lot of space
- can’t be rack mounted
- don’t often come with redundant power supplies
- cable organization is challenging
best uses for tower servers
- smaller offices
- central IT office that preconfigures servers to be used in branch offices
MDFs
- main distribution frames
- cable racks that organize data cabling from IDFs
IDFs
- intermediary distribution frames
- organize cables plugged into equipment
rack-mounted equipment
- increases the potential server density in a server room or data center
- most racks have front/back doors that can be locked (increases security)
UID
- unit identification
- used to identify servers in rack
- LED on front/back of rack-mounted server will blink when the server is being remotely managed or when the UID button is pressed on the server
rack-mounted servers
- thin computers designed to be stacked vertically in a rack
- provides organization
- uses minimal space (scalable)
- normally have dual power supplies
- normally have at least 2 network cards/management ports
- all connections on back of device
- ports on front of server for attaching peripherals
racks
- server room/datacenter must be able to handle weight of racks
- should be balanced and bolted to floor
- most are 19 in wide
- come with rails for mounting equipment or rail kits are available
- good to place storage appliances, power sources, etc. in rack with servers
U
- rack units
- unit of measurement for height
- vertical distance between the holes in the rack to which rails/rack-mounted equipment are secured
- 1U up to 7U per rack-mounted device
- most racks have a max of 42U
1U
1 3/4 inches
2U
3 1/2 inches
4U
7 inches
PDUs
- power distribution units
- provide power outlets to racks in server rooms/data centers
- redundant PDUs should be plugged into separate circuits
- redundant server power supplies should plug into separate PDUs
- check PDU rating to ensure equipment doesn’t draw more power than the PDU’s load capacity
cable management arm
- metal/plastic folding component that is attached to the back of a rack-mounted device
- all cables from the device are placed into the arm
- arm expands when device is pulled from rack to avoid pulling out cables
- prevent blockage of fan intake
- zip tie cables together from a device
rack fillers
- blanking panels
- cover empty spaces in the rack
- prevent blockage of fan intake
- measured via the U system
- available with venting holes
blade servers
- circuit board containing its own processors, memory and network capabilities
- may include a small amount of storage
- no power supply
- no cooling mechanisms
- can’t run independently
- most have USB connector on front
advantages of blade servers
- increase server density (scalability)
- will likely replace rack-mounted servers
disadvantages of blade servers
- more expensive
- harder to expand
blade enclosure
- proprietary chassis that can house several blade servers from the same vendor
- measure from 6U to 12U
blade enclosures provide
- temperature control mechanisms i.e. fans
- power
- network connectivity
- storage connectivity
- server remote management connections
backplane
- connects server and I/O blades within enclosure
- some provide redundancy
I/O blades
- cards
- provide faster network connectivity
- provide storage for blade servers
- provide management capabilities
midplane
- printed circuit board (PCB)
- server blades that connect on front
- other components accessible on back
- different components may require a specific midplane
- some provide redundancy
firmware
- software stored in a chip
- needs to be updated periodically
- updating firmware may resolve OS instability
- be sure updating to correct version
UUID
- unique 128-bit universal unique identifier in firmware
- may be updated when updating motherboard firmware
MD5/SHA hash values
- provided to confirm that firmware downloaded from internet has not been changed
BIOS
- basic input-output system
- firmware built into circuit board or RAID disk controller
- checks critical hardware components when booting
- power on self test (POST)
- checks master boot record (MBR) to hand control over to OS
- configuring BIOS depends on manufacturer
disks are initialized on a computer as either:
- MBR
- GUID partition table (GPT)
- only 1 copy of the MBR on an MBR disk (sits on the 1st sector of the disk before the 1st partition)
- GPT stores multiple copies of the MBR throughout the disk for additional resiliency
max partitions MBR disk
4
max partitions GPT disks
- 128 on a Windows system
- booting from a GPT disk is possible as part of the POST on unified extensible firmware interface (UEFI) system
CMOS
- complementary metal oxide semiconductor
- specific configuration of hardware settings supported by the BIOS
UEFI
- unified extensible firmware interface
- new BIOS
- GUI with mouse support
- GPT hard-disk support (including booting from GPT)
- IPv6 support during boot
- support for secure boot
- support for new modules for additional functionality
secure boot
- prevents untrusted executable code from running before OS initializes
- digital signatures used to validate trusted code
some Linux distros don’t support secure boot
must disable UEFI Secure Boot from UEFI settings to install these distros
CPU
- central processing unit
- brain of server
- 1st version = step 0
- step value increases as manufacturer releases the same CPU with improvements
scaling up
adding more processing power to a server
scaling out
adding more servers
CPU architecture
- most modern CPUs have 64-bit data path
- may also see 32-bit
- newer server OSs support only 64-bit
max addressable memory limit 32-bit machines
4GB
max addressable memory limit of 64-bit machines
16EB (exabytes)
ARM processors
- advanced RICS machine processors
- used in modern mobile devices
- available in 64-bit or 32-bit
CPUs and virtualization
- CPU must support hardware virtualization
- AMD processors must support AMD virtualization (AMD-V)
- Intel processors must support Intel virtualization technology (Intel VT)
buses
- move data among various components
- every motherboard always includes several buses
- bus speed (MHz) is different from internal CPU speed
- determines how quickly data moves between system components
- used for expansion cards and embedded components
CPU cache
- CPUs have small amounts of high-speed memory to speed up execution of anticipated CPU instructions (hit rate)
- hit rate indicates that CPU retrieved data/instructions from cache (hit) instead of from a slower location
- L1, L2, L3 caches common in modern server hardware
- L1, L2, L3 caches are typically built into the processor
- L4 cache uses main system memory
- common CPU cache sizes vary between CPUs and motherboards
- L3 and L4 cache memory is used by all CPU cores for CPU chips in a socket
L1 cache
- fastest
- less common
L4 cache
- slowest
- largest
why L1 cache is faster than normal system memory
- uses static RAM (SRAM)
- instead of dynamic RAM (DRAM)/synchronous DRAM (SDRAM)
SRAM
- uses transistors to retain data
- speeds up data access
common L1 cache size
96KB
common L2 cache size
2MB
common L3 cache size
8MB
common L4 cache size
16MB
LRU caching algorithm
- least recently used
- ensures cached items are listed in the order in which they were processed
- oldest items removed from cache making space for new items
MRU caching algorithm
- most recently used
- newest items in cache are removed when space is needed
SMP
- symmetric multiprocessing
- server can have multiple identical physical CPUs working together
- server motherboard must have sockets to accommodate the CPUs
- reduces time required to complete processing work
RISC
- reduced instruction set computing
- processors designed to work with simpler/smaller instruction sets
- more efficient
- smaller chip size than complex instruction set computing (CISC) processors
- often found in mobile devices
multiple CPU cores
- multiple CPUs within the same chip connect to a single motherboard socket
- uses less space than SMP
how SMP and multicore systems present CPU data
OS monitoring tools i.e. task manager
GPUs
- graphics processing units
- specialized processors designed to handle computationally intensive graphic operations
- freeing up CPU to complete other tasks
- can be embedded within the motherboard as part of the chip set
- available as expansion video cards
- can have dedicated video memory to isolate graphic from nongraphic operations
- server may require GPU depending on work (more common in high-end PCs)
cloud computing and GPU capabilities
- users may experience added benefits
- incurs greater usage cost
RAM
- random access memory
- electronic volatile memory used by running programs
- needs electricity to retain data
- servers can accommodate different amounts/types of RAM
- motherboards normally require memory chips to be installed in pairs for efficient use by CPU sockets
memory timing
- schemes can transmit data on the rise/fall of a clocking (timing) signal
- double pumping
- data transfer rate is effectively doubled
factors influencing memory speed
- time to select a memory column
- time to select a memory row
- time to read data when memory row is selected
ECC
- error correcting code memory
- can be accommodated by most server motherboards
- memory chip with ability to detect/fix memory errors at the bit level
- uses parity bit for each group of memory bits
- more expensive
server enables use of both ECC and non-ECC memory
- pair ECC chips together
- pair non-ECC chips together
- otherwise memory error correction will be disabled
DDR
- double data rate memory
- increases data transfer rates to/from memory
- compared to single data rate (SDR)
- different DDR versions that are not compatible with each other
- DDR3 chip sockets on motherboard will only accept DDR3 memory chips
DDR version performance
- higher number = best performance
- DDR1 < DDR2 < DDR3 < DDR4
- choose DDR4 if motherboard supports DDR2 - DDR4
CAS latency
- column access strobe
- defines how long it takes for requested memory contents to leave the chip on way to bus
HBA cards
- host bus adapter
- enables use of additional buses not supported by the motherboard
- SAN connectivity
- slimmer form factors benefit from riser circuit boards that plug into the motherboard and enable expansion cards
PCI
- peripheral component interconnect
- 32-bit and 64-bit bus and expansion slot card standard
- legacy
- 64-bit PCI cards could be plugged into 32-bit PCI expansion slots
- full-height PCI cards may not fit form factor
- rated between 133 - 800 MBps
PCI-X
- PCI extended
- 64-bit parallel transmission standard
- provides transfer rates up to 4x to/from slots
- bit transfer between 2 - 4 GBps
- runs at various frequencies (influences data transmission speed)
PCIe
- PCI express
- serial-transmission based standard
- supersedes PCI/PCI-X with more bandwidth
- video/storage/network cards
- bandwidth varies based on specific version of PCIe and how many lanes (channels of communication) are used
- opt for x16 (16 lane) PCIe cards if slots are compatible for best performance
- supported by most modern servers
- may not be hot-pluggable
- can plug an x8 card into a x16 slot
DMA
- direct memory access
- supported by some PCIe devices
- enables access to system memory without involving CPU
NICs
- network interface cards
- most physical servers have at least 2 integrated
- virtual servers can have as many virtual NICs (vNICs) as required for the server role
- vNICs can be tied to specific underlying physical NICs
- each NIC needs to be configured with the appropriate IPv4/IPv6 settings
adding NICs to physical server
- need to know which type of unused expansion slots are in motherboard
- get fasted NIC that will work in environment
storage options for VMs running server OS
- use physical disk partition available from underlying hypervisor
- use storage accessible over a SAN through underling hypervisor
- use a virtual hard disk (VHD) file
- VHD files may be stored on a local disk on underlying server or on a network storage location
physical server storage options
- direct-attached storage (DAS)
- network-attached storage (NAS)
- storage area networks (SANs)
- expansion card connected to server for additional storage
RAID controllers
- most server motherboards include an integrated RAID controller that connects to multiple physical disks
- have their own firmware configuration utility
USB
- universal serial bus
- used for a wide variety of components/peripherals
- servers normally use for keyboards and mice (HIDs)
- USB 1.0 - 4.0
- USB 3.0 is backward-compatible
HID
human interface device
KVM switches
- used in server rooms/data centers
- link numerous servers to a single keyboard, display an mouse
- connections are made via USB or IP addresses on network
power connections
- 2 prong
- 3 prong with ground
- NEMA/Edison connectors
- industrial grade equipment uses twist-lock mechanism to prevent accidental disconnect
voltage
- CPUs don’t need as much power as the amount often fed from power supplies
- voltage regulator module (VRM) ensures CPU only gets the voltage it needs
- normally built into motherboard
control electrostatic discharge
- connect ESD strap (bracelet) alligator clip to computer chassis
- stand on ESD mat while servicing hardware
- ESD footwear
- touch computer chassis as often as possible to dissipate any built up charge
- store components in ESD bags
wattage
- server components draw power from power supply unit (PSU)
- most server-class hardware uses redundant PSUs
- calculate sum of power required for all internal components to ensure PSU form factor matches server chassis form factor
- PSUs have total wattage listings
UPS
- uninterruptible power supply
- batteries are constantly charged from the standard electricity power grid
- servers connected to UPS can at least shut down gracefully during power outage
- normally rack-mounted in server rooms/data centers
- UPS runtime capacities specify how long UPS batteries can support a specific wattage
standby UPS systems
quickly switch to battery power when main power source fails
online UPS systems
- always use batteries
- batteries constantly being refreshed when main power source is available
- ensure consistent regulated flow of electricity
- more expensive
- appropriate for minimizing disruptions to mission critical IT services
calculating UPS power draw
- number of servers in rack
- power draw per server
- power draw per storage array
- how busy each server is
UPS monitoring/management
- conduct testing periodically to ensure batteries will provide power when needed
- use bypass devices to ensure UPS power load is not interrupted when maintenance is required
- don’t plug non-essential devices into UPS
- UPS can be connected to the server via USB cable
- UPS can be plugged into the network
- remote monitoring/management is possible
- most vendors offer a mobile app for UPS monitoring
environmental controls
- temperature
- airflow (feed in cool air)
- separate hot air and cold air via hot and cold aisles
- humidity (too little increases likelihood of ESD)
- RH should fall between 40 - 60%
- fire suppression (avoid water sprinklers if possible)
L1 cache size
96KB
L2 cache size
2MB
L3 cache size
8MB
l4 cache size
16MB
CPU type for LGA 1248 socket
Intel Itanium 9300
CPU type for LGA 1567 socket
Intel Xeon 6500/7500
CPU type for LGA 2011 (Socket R)
- Intel Core i7
- Intel Xeon E5
CPU type for LGA 1155 (Socket H2)
Intel Xeon E3
DDR2 physical pins/bus speed
- 240
- 200 - 500 MHz
DDR3 physical pins/bus speed
- 240
- 400 - 1000 MHz
DDR4 physical pins/bus speed
- 288
- 1000 - 2000 MHz
PCIe 1.0 data transfer rates
- 1 channel = 250 Mbps
- 2 channels = 500 Mbps
- 16 channels = 4 Gbps
PCIe 2.0 data transfer rates
- 1 channel = 500 Mbps
- 16 channels = 8 Gbps
PCIe 3.0 data transfer rates
- 8 channels = 8 Gbps
- 16 channels = 15 Gbps
USB 1.0 data transfer rate
12 Mbps
USB 2.0 data transfer rate
480 Mbps
USB 3.0 data transfer rate
5 Gbps
USB 4.0 data transfer rate
40 Gbps
class A fire extinguisher rating
- paper
- wood
- cloth
class B fire extinguisher rating
- solvents
- propane
- gasoline
class C fire extinguisher rating
electrical equipment
class D fire extinguisher rating
combustible metals