diff options
| -rw-r--r-- | unit5/assets/bcwn.png | bin | 0 -> 51660 bytes | |||
| -rw-r--r-- | unit5/assets/cachetier.png | bin | 0 -> 22488 bytes | |||
| -rw-r--r-- | unit5/assets/virtstorpro.png | bin | 0 -> 80188 bytes | |||
| -rw-r--r-- | unit5/unit5.typ | 74 |
4 files changed, 74 insertions, 0 deletions
diff --git a/unit5/assets/bcwn.png b/unit5/assets/bcwn.png Binary files differnew file mode 100644 index 0000000..da1fef8 --- /dev/null +++ b/unit5/assets/bcwn.png diff --git a/unit5/assets/cachetier.png b/unit5/assets/cachetier.png Binary files differnew file mode 100644 index 0000000..9e15a95 --- /dev/null +++ b/unit5/assets/cachetier.png diff --git a/unit5/assets/virtstorpro.png b/unit5/assets/virtstorpro.png Binary files differnew file mode 100644 index 0000000..ef4c881 --- /dev/null +++ b/unit5/assets/virtstorpro.png diff --git a/unit5/unit5.typ b/unit5/unit5.typ index 069a164..0844ea1 100644 --- a/unit5/unit5.typ +++ b/unit5/unit5.typ @@ -233,3 +233,77 @@ _Resource Management is the process of allocating resources effectivly to a serv - Copy of the hottest data resides on the flash cache. - Server flash cache needs warm up time before performance increase is observed. - Warm up time is the time required to move significant amount of data into server flash cache. +== Storage +=== Virtual Storage Provisioning +#figure(image("./assets/virtstorpro.png", width: 50%)) +_Virtual storage provisioning is the process that enables the presentation of a LUN to an application with more storage than is physically allocated to it on the storage system._\ +- Administrators often over-anticipate storage requirements. +- This leads to unused space and lower capacity utilization. +- It also leads to excess storage capacity, which begets higher cost, increased power consumption, cooling and floor space. +- With virtual storage provisioning, physical storage is allocated on demand. +- More efficient storage utilization by reducing allocated but unused storage. +- Provides rapid elasticity by adapting to variations in workloads by dynamically expanding or reducing storage levels. +=== Storage pool rebalancing +_Storage pool rebalancing is a technique that allows to automatically rebalance allocated extents on physical disk drives over the entire pool when new drives are added to the pool._\ +- When storage pool is expanded, sudden introduction of empty drives combined with old full drives causes data imbalance. +- Also causes performance impact because new data would be added mostly to new drives. +- Restripes data across all drives in the storage pool. +- This enables spreading out data equally on all physical disks within the shared pool. +- This ensures that used capacity of each drive is uniform across the pool and helps in increasing overall pool performance. +=== Storage space reclamation +- Identifies unused space in thin LUNs and returns it to the storage pools. +- Two options for storage reclamation are: + 1. *Zero extent reclamation*\ + - Commonly implemented at the storage level. + - Provides the ability to free storage extents that only contain zeros. + 2. *API based reclamation*\ + - Uses APIs to communicate locations of unused space in LUNs to storage system. + - Allows storage system to reclaim all unsued physical storage back into the storage pools. +=== Automated storage tiering +_Automated storage tiering is the technique of establishing a hierarchy of different storage types for different categories of data that enables storing the right data automatically to the right tier, to meet the service level requirements._\ +- Many services have predictable spikes in activity with lower activity at other times. +- Automated storage solution addresses these cyclical fluctuations as well as unpredictable spikes. +- It can replace manul storage management and significantly benefit cloud environments. +- Tiers are differentiated based on protection, performance, cost. +- Example, using tier 0 SSDs to store hot data and tier 1 HHDs to store cold data. +- Optimizes complete use of all kinds of storage. +- Data is moved between tiers based on defined tiering policies. +- These policies are based on file type, frequency of access, etc. +- Data movement between tiers can happen within or between storage arrays. +=== Cache tiering +#figure(image("./assets/cachetier.png", width: 50%)) +- Tiering can also happen at the cache tier. +- A large cache improves performance by retaining large amount of frequently accessed data. +- High proportion of reads happen from the cache. +- Configuring large cache can be costly. +- The size of cache can be increased by using SSDs on the storage system to create a large capacity secondary cache seperate from compute's primary cache. +- This enables tiering between DRAM primary cache and SSDs secondary cache. +- Enables storage system to store a lot of hot data on the cache tier. +- Most reads can now happen from cache tier increasing performance. +=== Dynamic VM load balancing across storage volumes +- During provisioning for VMs, volumes are selected randomly. +- This can lead to underutilized volumes. +- Dynamic VM load balancing across storage volumes enables intelligent placement of VMs during creation. +- It does so based on I/O load, available storage in hypervisor's native FS or NAS FS. +- Implementd in a centralized management server the manages virtualized environments. +- The server performs ongoing dynamic VM load balancing within a cluster of volumes. +_A cluster volume is a collection of pool of a hypervisor's native FS or NAS FS volumes that are aggregated as a single volume to enable efficient and rapid placement of new virtual machines._\ +- User configurable space utilization and I/O latency thresholds are defined to ensure space efficiency. +- I/O bottlenecks avoided. +- Thresholds configured during configuration of clustering volumes. +== Network +_Network traffic flow in a cloud network infrastructure is controlled to optimize both performance and availability of cloud service._\ +- Administrators use several traffic management techniques. +- Some enable distribution of traffic load across nodes or parallel network links. +- This prevents overutilization and underutilization of resources. +- Othes enable automatic failover of network traffic from a failed network component. +- Some also ensure garenteed service levels for a class of traffic contending with other classes for network bandwidth. +=== Balancing client workload across nodes +#figure(image("./assets/bcwn.png")) + +=== Network storm control +=== Quality of Service (QoS) +=== Traffic shaping +=== Link aggregation +=== NIC Teaming +=== Multipathing |
