Advantages Of Flat Address Space, Role Of Metadata Service, Memory Virtualization, RAID, Storage Design, And Storage Evolution

Flat Address Space vs Hierarchical Address Space

The process of Memory virtualization worked with Virtual Machine, Process and Memory Address (García-Valls, Cucinotta & Lu, 2014). When multiple Virtual Memory (VM) is connected in series and a process (P) is taking place in each VM then the mapping is done in the memory inside one-page table for cache mechanism to get easy access in the Memory Address (MA). The two components to run this addresses are Translation Lookaside Buffer (TLB) and Memory Management Unit (MMU) that can optimize all the performance that are in the virtual memory (Gandhi et al., 2014). Mapping of actual memory with the Guest machine memory is done by the mechanism called shadow paging, which comes from the Database Management System (DBMS). This whole mechanism of Memory Virtualization works internally with hexadecimal codes, which is managed by the Operating System.

RAID means (Redundant Array of Independent Disks), RAID 0 is for the configuration used striping for data handling (Fekete & Varki 2016). It increases the performance of the systems. In RAID 0 all the data files are first broken into smaller blocks each and written to a separate disk drive to increase the I/O performance with little to no fault tolerance.

The term mirroring, parity and striping are the storage techniques of RAID:

Stripping is a process in which data files in a RAID are broken down into smaller block size each and written to a different physical disk drive.

Mirroring is a technique to store the data of identical copies on the RAID simultaneously. This affects fault tolerance and the performance.

Parity is a technique where striping and the checksum method is used for the data blocks to calculate the parity function.

Exercise 2: Storage Design (2 marks)

Storage Repository Design, the other name known is LUN (Logical Unit Number) in the industry of IT(Ume, 2015). Storage Repository Design has developed their importance for snapshot design, usage of the disk and for the cache and the performance of the usage of disk. SAN allocates cache to each individual repository containing the write cache. In a snapshot, create a storage repository, which is a delta file, of exactly the same size of the Virtual Disk within the storage repository and allow quick easy rollback. One disadvantage is, if the Virtual disk is lost, then the snapshot cannot be rolled back as delta disk. Underlying the performance can only be achieved in the matter of physical disk is

Role of Metadata Service in OSD Node

The Internet Storage System (ISS) referred as the RAID arrays that arise feature-rich for the storage arrays and are capable of providing highly optimized I/O processing (Aderholdt et al., 2015).

1. Front End: The Front End interfaces the host and a Computer System or the Storage System. The Front End has mainly two components-
   1. Front End Port: Front End Ports help to connect the host to the intelligence storage system.
   2. Front End Controllers: Front End Controllers control the data coming from the internal Data bus.
2. Cache is a semiconductor.
3. Back End interfaces the Physical disks and the Cache Memory. It has two components:-
   1. Back End Ports: The Back End ports controls the transference of the data between the physical disks and the Cache memory
   2. Back End Controllers: Back End controller will communicate with the particular disks to Read/Write the operation to provide some limited data storage temporarily.
4. Physical Disks: In the Physical Disks all the data is store and are then connected with the back end with the help of Fiber Channel Interface.

1. Cache main purpose is to reduce the time for the I/O service to request the host (Bender, 2016). The host is isolated from only the mechanical delay, which is connected to the hard disk drive or to the rotating disk so that the performance of the Storage System can be improves by Cache. For ISS, the rotating disk is the slowest component. 

SAN is a block I/O disk directly connected to the server. NAS is a remote I/O file, which is redirected from a network to a device.

SAN advantages over NAS are as follows (Patil, 2016):

• SAN is faster, scalable and simple to use than NAS system.
• NAS runs devices on a built-in OS, whereas SAN uses Fiber Channel to connect via storage devices and make it capable of sharing data of low level among themselves.

The two known NAS file sharing protocols are (Maciel, 2014):

NFS – Network File System Protocol, which is define as a protocol to share a file of the Traditional UNIX environment.

CIFS – The Protocol of Common Internet File System, which depends on the Server Message Block Protocol for Traditional Microsoft environment file sharing protocol

1. Solution:

Time required for I/O depends on disk service time

Therefore, Disk service time = Average seek time + data transfer time + rotational latency

Given, Average seek time = 5 millisecond

Rotational latency = ½ time taken for the full rotation.

Since, rotational speed = 15000 rpm

1 revolution = 1/ (15000/60) sec.

Therefore, For ½ revolution time taken = 0.5/ (15000/60) = 2ms.

Data transfer rate = 40 MB/s

Therefore, 4 KB I/O transfer takes, 4 KB/40MB/s= 0.1 ms

Therefore, Time required for 1 I/O to perform = 5 ms + 2ms + 0.1ms= 7.1 msec

Now, a disk can perform maximum number of IOPS = 1/7.1 ms = 140 IOPS

Therefore, the maximum number of IOPS a disk perform at 70 % utilization = 140 * 0.7 = 98 IOPS

Now, to meet the disks number,

The performance of Application’s required = 4900/98 = 50 disks

The capacity of the Application’s required = 1 TB/100 GB = 10 disks

Finally, the disk that is required = Maximum (Capacity, Performance)

= Maximum (10, 50) = 50 disks

Storage Evolution (2 Marks)

Fibre Channel over Ethernet (FCoE) provides simplicity, scalability to the Ethernet with lossless nature, functionality, manageability and offers huge savings in reducing the number of an adapter and the switches. It allows the user to leverage with the existing investment in Information Technology, lowers the cost and improves the performance.

FCoE replaces the NIC and the HBA in the server and then uses connections of 10 GB to provide losslessness to the FCoE switch

Virtual Storage Area Network (SAN) defines the configuration system, which will be easily scaled out easily without changing the physical layout.

The protocols of IP SAN are those protocols having the storage devices with multiple servers of which has been shared block that can access pools by using storage protocol, which depends only on the Internet Protocol (IP) suite (González-Férez & Bilas, 2014). Fibre Channel over IP is a storage networking technology that is completely based on the Internet Protocol (IP), and the Internet Engineering Task Force (IETF) has developed it (Black & Koning, 2014)that enables all the transmission that are done for the Fibre Channel (FC) for tunneling the data between the SAN over the IP network.

Memory Virtualization and Performance Optimization

What is an advantage of a flat address space over a hierarchical address space?

Highly scalable with minimal impact on performance

Provides access to data, based on retention policies

Provides access to block, file, and object with the same interface

Consumes less bandwidth on network while accessing data

Answer: a. Highly scalable with minimal impact on performance

What is a role of metadata service in an OSD node?

Responsible for storing data in the form of objects

Stores unique IDs generated for objects

Stores both objects and objects IDs

Controls functioning of storage devices

Answer: b. Stores unique Ids generated for objects

What is used to generate an object ID in a CAS system?

File metadata

Source and destination address

Binary representation of data

File system type and ownership

Answer: c: Binary representation of data

What accurately describes block I/O access in a unified storage?

I/O traverse NAS head and storage controller to disk

I/O traverse OSD node and storage controller to disk

I/O traverse storage controller to disk

I/O is directly sent to the disk

Answer: c: I/O traverse storage controller to disk

What accurately describes unified storage?

Provides block, file, and object-based access within one platform

Provides block and file storage access using objects

Supports block and file access using flat address space

Specialized storage device purposely built in archiving

Answer: a: Provides block, file and object-based access within one platform

Reference

Aderholdt, F., Caldwell, B. A., Hicks, S. E., Koch, S. M., Naughton III, T. J., Pogge, J. R., … & Sorrillo, L. (2015). Secure Storage Architectures (No. ORNL/TM-2015/212). Oak Ridge National Laboratory (ORNL).

Bender, M. A., Demaine, E. D., Ebrahimi, R., Fineman, J. T., Johnson, R., Lincoln, A., … & McCauley, S. (2016, July). Cache-adaptive analysis. In Proceedings of the 28th ACM Symposium on Parallelism in Algorithms and Architectures(pp. 135-144). ACM.

Black, D., & Koning, P. (2014). Securing Block Storage Protocols over IP: RFC 3723 Requirements Update for IPsec v3 (No. RFC 7146).

Fekete, A., & Varki, E. (2016, September). RAIDX: RAID without striping. In Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS), 2016 IEEE 24th International Symposium on (pp. 418-423). IEEE.

Gandhi, J., Basu, A., Hill, M. D., & Swift, M. M. (2014, December). Efficient memory virtualization: Reducing dimensionality of nested page walks. In Proceedings of the 47th Annual IEEE/ACM International Symposium on Microarchitecture (pp. 178-189). IEEE Computer Society.

García-Valls, M., Cucinotta, T., & Lu, C. (2014). Challenges in real-time virtualization and predictable cloud computing. Journal of Systems Architecture, 60(9), 726-740.

González-Férez, P., & Bilas, A. (2014, June). Tyche: An efficient Ethernet-based protocol for converged networked storage. In Mass Storage Systems and Technologies (MSST), 2014 30th Symposium on (pp. 1-11). IEEE.’

Maciel, P., Matos, R., Callou, G., Silva, B., Barreto, D., Araujo, J., … & Worth, S. (2014, October). Performance evaluation of sheepdog distributed storage system. In Systems, Man and Cybernetics (SMC), 2014 IEEE International Conference on(pp. 3370-3375). IEEE.

Patil, P. T. (2016). A Study on Evolution of Storage Infrastructure. International Journal, 6(7).

Ume, L. E. (2015). Effective Data Backup System Using Storage Area Network Solution. West African Journal of Industrial and Academic Research, 14(1), 18-27.