PowerMax and VMAX All Flash Solutions Design (Fibre Channel Considerations) Flashcards
What is required for FCP data path?
two round trips for each write
How do writes work w/ FCP?
initatior issues write command
receiver initiates transfer ready
initiator sends data
receiver sends status good
Why is FC still fast even though it needs two round trips per write?
normal host to SAN operations are short distance
latency of two writes is negligible
What is the latency of FC replication?
1ms latency per 200km
4ms latency per write for sites 200km away
What is Write Acceleration/Fast Write?
performance enhancement feature offered by some FC products
goal is to minimize latency caused by the two round trips - increases performance for synchronous replication
How is Write Acceleration/Fast Write initiated?
intelligent switches have additional buffers and switches at either end for replication have to coordinate operations
What is the latency of Write Acceleration/Fast Write?
equivalent of one round trip
What effect does Write Acceleration/Fast Write have on SRDF/S?
will always increase performance of SRDF/S
What effect does Write Acceleration/Fast Write have on SRDF/AR or adaptive copy?
improves SRDF throughput by 20%
What effect does Write Acceleration/Fast Write have on SRDF/A?
not a lot of performance increase since SRDF/A sends a continous stream of writes until the cycle is empty
if links are kept full there is no waiting for round-trip delay and SRDF/A performance not impacted by two round trips
When will Write Acceleration/Fast Write increase SRDF/A performance?
when there are gaps on the links
due to retransmissions after link errors
or small writes that throttle SRDF throughput due to hitting IO per second limit on links
What are best practices for implementing FC in PowerMax/VMAX environments?
use Dell EMC SiRT or 3rd party Fast Write feature
don’t use both simultaneously
connect host and storage pairs to same switch to reduce latency
What does FC use for flow control?
BB_Credit
How does BB_Credit work?
providing for frame acknowledge via R_RDY primitive defined in the FC1 Layer
BB_Credit management may occur between any two FC ports that are connected
What are the two most common BB_Credit connections?
between an N Port and F Port
Between two E Ports
What is required for transmitting ports to send frames?
only if BB_Credits are greater than 0
an R_RDY is sent from the receiving port to the transmitting port for every available buffer on the receiving side
What is the relationship between BB_Credits and long distance replication?
insufficent BB_Credits will result in the degradation of throughput over long distance links
longer the link the more frames must be sent continously down the link to prevent to prevent this degradation
How are BB_Credits sized in FC connections?
sized for 100% utilization on long distance links
switches at either end must support the minimum required level of BB_Credits
recommended to add 20% buffer to account for traffic spikes
How are BB_Credits calculated on a switch?
on a per port basis
What are the methods to extend FC connections?
long wave laser optics w/ ISLs
FC traffic over CWDM/DWDM/SONET
FCIP
What is fabric isolation?
goal is to define fabrics in a way that events occuring in one fabric will not impact other fabrics
Why use fabric isolation with FC based replication?
isolating traffic locally while minimizing traffic across the WAN leads to higher performing remote replication
How can fabric isolation be achieved?
using dedicated switches to form the link across the WAN
What is a typical scenario for FC remote replication fabric isolation?
each site will have its operation fabric where hosts access storage
a third fabric is created for traffic across the WAN - not connected to local/remote fabrics
disadvantage of this method is inability to share resources between local and remote sites
What is a FC routing method of FC remote replication?
connects two physically separate fabrics to enable shared access to storage resources from any fabric
only the shared resources will be routed across the WAN
What is FCIP?
tunneling protocol that allows geographically distributed FC SANs to be connected over local IP
each interconnection is called a FCIP Link
How does FCIP work?
no translation of FCIDs
frames encapsulated into IP payloads
What is the size of FC frames?
typically 2148 bytes w/ MTU size of 1500
can get to MTU 9000 but MTU must be set on local and remote ports the same
What is TCP/IP used for in FCIP?
underlying transport to provide congestion control and in-order delivery of error-free data
What is ISL Trunking?
allows multiple ISLs to function as a single ISL
aggregates bandwidth of all the ISLs
achieves dynamic load balancing between ISLs
When should you implement ISL Trunking for FC replication?
when the (# of remote ports connected to a switch X port speed) + other traffic > (# of ISLs X ISL Speed)
when any ISL is observed to be over-utilized while an adjacent ISL is minimally utilized
What is Exchange-Based Trunking?
available on Brocade and QLogic switches
works by distributing exchanges evenly over a group of ISLs
What is Open Trunking?
works by making periodic changes to the hardware routing tables based on amount of time being spent at zero BB transmit credit and percent utilization measurement
What is Frame Based Trunking?
available on B Series Brocade
involves aggregation of several ISLs between any two adjacent switches into one logical unit so FC frames can be evenly distributed across ISLs
