====== Selecting Cloud service Provider ====== {{:ddvsaws.jpg|}} ===== Hardware Configuration ===== Haproxy - 2 vCPU, 4GB ram, 10GB HDD (centos 6.x) App server x2 - 8 vCPU, 16GB ram, 150GB HDD (centos 6.x) - C4 - c4.2xlarge (EBS) / C3 - C3.2xlarge (SSD) prodcollab x2 na1/na3 - 6/8 vCPU, 16GB ram, 150GB HDD (Ubuntu 14.04) - C4 - c4.2xlarge (EBS) / C3 - C3.2xlarge (SSD) mysql x2 Master/Slave - 8 vcpu, 20GbM/16Gbs ram, 150GB HDD (centos 6.5) - c3.4xlarge/c3.2xlarge Storage (HDFS) - i2.xlarge, i2.2xlarge ===== Dimension Data (NA1 and NA3 sites) ===== Santa Clara speed test 192.168.1.10 (santa clara Test machines) - 18Mb/s Download speed - ~2MB/s Check URL : https://na3.learnexa.com/system/409-maya.mp4 - 2MB - Full speed reached however, we are left with 1Gb/s = 128MB/s (In Theory). However on NA3 server we tested for speed and here are the results (Speedtest.net public servers 1gbps link : maximumvps.net) Average Upload speed From clients to App server (NA3 - Prodapp01) - 5.1 MB/s Average Download speed From clients to App server (NA3 - Prodapp01) - 4.0 MB/s ==== Findings on Dimension Data ==== We have tested the environment and we found following issues; The only difference between NA1 and NA3 site is the high performance Disk drives (on NA3). Other issues are as follows; 1. Vendor lock-in, Each site is considered external network as there are no working data migration option. 2. No option to increase the Network speed stuck to 1Gbps on internal network. 3. NAT or Public network does not provide optimum performance, as mentioned above. 4. Since we are on single storage server using MAPR-FS (HDFS) does not really matter and what matters is the Network performance. ===== Storage performance ===== We used a 400MB unstructured data (Video) for the test. GlusterFS on XFS on NA3 818121492 100% 70.12MB/s 0:00:11 (xfer#1, to-check=0/1) GlusterFS on EXT4 on NA3 818121492 100% 68.62MB/s 0:00:11 (xfer#1, to-check=0/1) MAPR-FS (HDFS) on NA3 818121492 100% 124.73MB/s 0:00:06 (xfer#1, to-check=0/1) ===== AWS - Amazon Web Services ===== We have gone through their offerings and following are the key features; 1. 10Gbps with low latency Network available 2. High Performance Disk (SSD - Solid State Disks) with high I/O performance. 3. Better service model and availability. Can switch data-centers with no vendor lock-in. 4. Switching Zones is easy and Region is also possible however, requires manual transfer of AMI (Amazon Machine Image). We need to choose the right machine from AWS instance matrix as mentioned below. https://aws.amazon.com/ec2/instance-types/ ==== Amazon Hourly Pricing (Last updated 25-SEP-2015) ==== {{:amazon_aws_hourly_price.png|}} === The options we selected were === c4.2xlarge (EBS) - $452.352/Month (24x7 - 31 days) +/- I/O request \\ c3.2xlarge (SSD) - $393.576/Month (24x7 - 31 days) +/- I/O request \\ c3.4xlarge (SSD) - $787.152/Month (24x7 - 31 days) +/- I/O request \\ i2.xlarge (SSD) - $757.392/Month (24x7 - 31 days) +/- I/O request \\ i2.2xlarge (SSD) - $1514.04/Month (24x7 - 31 days) +/- I/O request \\ === Google Cloud === {{:google_cloud_pricing.png|}} For more information and calculating the estimates Please visit the following Website : [[https://cloud.google.com/products/calculator/]]