Update scalability-overview.md

Author: abinav2307

articles/cosmos-db/mongodb/vcore/scalability-overview.md

@@ -17,7 +17,7 @@ Vertical scaling offers the following benefits:
 - Application teams may not always have a clear path to logically shard their data. Moreover, logical sharding is defined per collection. In a dataset with several unsharded collections, data modeling to partition the data can quickly become tedious. Simply scaling up the cluster can circumvent the need for logical sharding while meeting the growing storage and compute needs of the application.
 - Vertical scaling does not require data rebalancing. The number of physical shards remains the same and only the capacity of the cluster is increased with no impact to the application.
 - Scaling up and down are zero down-time operations with no disruptions to the service. No application changes are needed and steady state operations can continue unperturbed.
-- Compute and Storage resources can also be scaled down during known time windows of low activity. Once again, scaling down avoids the need to rebalance data across a fewer number of physical shards and is a zero down-time operation with no disruption to the service. Here too, no application changes are needed after scaling down the cluster.
+- Compute and Storage resources can also be scaled down during known time windows of low activity. Once again, scaling down avoids the need to rebalance data across fewer physical shards and is a zero down-time operation with no disruption to the service. Here too, no application changes are needed after scaling down the cluster.
 - Most importantly, Compute and Storage can be scaled independently. If more cores and memory are needed, the disk SKU can be left as is and the cluster tier can be scaled up. Equally, if more storage and IOPS are needed, the cluster tier can be left as is and the Storage SKU can be scaled up independently. If needed, both Compute and Storage can be scaled independently to optimize for each component's requirements individually, without either component's elasticity requirements affecting the other.
 
 
@@ -43,7 +43,7 @@ Disk IOPS influences write operations in the vCore based service for Azure Cosmo
 As mentioned earlier, storage and compute resources are decoupled for billing and provisioning. While they function as a cohesive unit, they can be scaled independently. The M30 cluster tier can have 32 TB disks provisioned. Similarly, the M200 cluster tier can have 32 GB disks provisioned to optimize for both storage and compute costs. 
 
 ## Lower TCO with large disks (32 TB and beyond)
-Typically, NoSQL databases with a vCore based model limit the storage per physical shard to 4 TB. The vCore based service for Azure Cosmos DB for MongoDB provides upto 8x that capacity with 32 TB disks and plans to expand to 64 TB and 128 TB disks per shard soon. For storage heavy workloads, a 4 TB storage capacity per physical shard will require a massive fleet of compute resources just to sustain the storage requirements of the workload. Compute is more expensive than storage and over provisioning compute due to capacity limits in a service can inflate costs rapidly. 
+Typically, NoSQL databases with a vCore based model limit the storage per physical shard to 4 TB. The vCore based service for Azure Cosmos DB for MongoDB provides upto 8x that capacity with 32 TB disks and plans to expand to 64 TB and 128 TB disks per shard soon. For storage heavy workloads, a 4 TB storage capacity per physical shard requires a massive fleet of compute resources just to sustain the storage requirements of the workload. Compute is more expensive than storage and over provisioning compute due to capacity limits in a service can inflate costs rapidly. 
 
 Let's consider a storage heavy workload with 200 TB of data. 
 | Storage size per shard      | Min shards needed to sustain 200 TB | 
@@ -52,7 +52,7 @@ Let's consider a storage heavy workload with 200 TB of data.
 | 32 TiB                      | 7                                   | 
 | 64 TiB (Coming soon)        | 4                                   | 
 
-The reduction in Compute requirements reduces sharply with larger disks. While more than the minimum number of physical shards may be needed sustain the throughput requirements of the workload, even doubling or tripling the number of shards will be more cost effective than a 50 shard cluster with smaller disks.
+The reduction in Compute requirements reduces sharply with larger disks. While more than the minimum number of physical shards may be needed sustain the throughput requirements of the workload, even doubling or tripling the number of shards are more cost effective than a 50 shard cluster with smaller disks.
 
 ## Skip storage tiering with large disks
 An immediate response to compute costs in storage heavy scenarios is to "tier" the data. Data in the transactional database is limited to the most frequently accessed "hot" data while the larger volume of "cold" data is detached to a cold store. This causes operational complexity. Performance is also unpredictable and dependent upon the data tier that is accessed. Furthermore, the availability of the entire system is dependent on the resiliency of both the hot and cold data stores combined. With large disks in the vCore service, there is no need for tiered storage as the cost of storage heavy workloads is minimized.