Sunteți pe pagina 1din 3

10 Rules for Scalable Performance in Simple Operation Datastores Questions & Answers -Sriharsha Uppalapati (A01646114)

Nare Hayrapetyan 1) What are the advantages of column-storedatabases and why are they not widely used? They are preferred when only particular columns are required. It is easier to retrieve the required data. The unused columns are left out. They are good for data warehouse. They would result in faster response time. It is not widely used as it is new and growing.

Naga Sai 1) Why shared nothing DBMS is better than SMP? SMP is limited by main memory. Difficult to synchronize.

Kamilia Ahmadi 1)In shared-nothing approach stated that if we have data skew or hot spot ,then the system degrade in performance? What is the reason for this behavior? All the data will accumulate at a hot spot. Lots of modifications of particular data.

Arvind Konda 2)How did they overcome synchronization problems faced by second option(collections of CPU with private main memory and share a common disk system) in the third option of using Shared nothing scalability.(Is it only by avoiding multi-shard operations?)

Due to the private buffer pool in each nodes main memory,the same disk block can be in multiple buffer pools.Hence,careful synchronization of these buffer- pool blocks is required. Shared-nothing systems scale only if data objects are partitioned across the systems nodes in a manner that balances load.

Rajasekhar Reddy 1)If updates are required within a transaction to multiple shreds then how the data consistency across the nodes is guaranteed by the application? Application logic will take care of the consistency. Even if the data is partitioned into multiple shreds, the transaction should happen on both. The lock has to be distributed across all nodes

Brett Francis 2) What are the reasons that the NoSQL movement (key-value stores,document stores and extensible record stores) is willing to abandon ACID(Atomicity,Consistency,Isolation,Durability) transactions? Performance-if it is a read database ,it does not need to perform transaction. For high availability,abandon consistency.

Arvind Konda 3)CAP Theorem. Elaborate and examples. Nawaf 6)what are the severe drawbacks of sharing in application logic? Hao 1) Can you give us some specific advices which will enhance CPUs percentage of usage on useful work(Based on this paper,only 13% of CPUs usage is used on useful work) Main memory databases are required to reduce CPU cycle usage on not so important tasks. Use existing databases. Hard to write code and maintain it. In the presence of certain failures,it states that a distributed system can have only two out of the three characteristics:consistency,availability and partition-tolerance. Recovery from disasters is important and should be viewed as an extension of high availability.

Sandeep Reddy 4) Why node performance should not be ignored while considering linear scalability. System works only as fast as the slowest node.

Naga Sai 1) Why node failures are most common as system scales? The more nodes we have, the probability of a node failure increases Replication will improve fault tolerance but at the cost of slowing the updates.

Kamilia Ahmadi 2)How using stored-procedures in Rule 2 can help in decreasing the overhead of communication? In case of stored procedures, a transaction is a single over-and-back message. The DBMS further reduces communication overhead by batching multiple transactions in one call.

Ashwin Kumar 3)There are customers who prefer a closed source over an open source product.So what are the factors or advantages of a closed source over an open source? Nawaf 4)Is there any order/priority we have to follow up for these 10 rules? Priority depends on customer requirements Depends on the context in which the Database server is being deployed. closed source-reliable and secure open source-no support

S-ar putea să vă placă și