A Dell EMC Best Practices Guide Dell EMC PowerEdge and BlueData Enabling Big-Data-as-a-Service and Exploratory Analytics on Dell EMC PowerEdge Compute Platforms, Ready Solutions, and Isilon Storage Kris Applegate ([email protected]) Dell EMC Customer Solution Center October 2017
25
Embed
Dell EMC PowerEdge and BlueData - NETWORLD€¦ · A Dell EMC Best Practices Guide Dell EMC PowerEdge and BlueData Enabling Big-Data-as-a-Service and Exploratory Analytics on Dell
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
A Dell EMC Best Practices Guide
Dell EMC PowerEdge and BlueData Enabling Big-Data-as-a-Service and Exploratory Analytics on Dell EMC PowerEdge Compute Platforms, Ready Solutions, and Isilon Storage
Kris Applegate ([email protected]) Dell EMC Customer Solution Center October 2017
Incorporated new 14th Generation of Dell EMC PowerEdge Servers
Added configurations for existing Hadoop and Isilon
Added design accommodations for new features in EPIC 3.0
The information in this publication is provided “as is.” Dell Inc. makes no representations or warranties of any kind with respect to the information in this
publication, and specifically disclaims implied warranties of merchantability or fitness for a particular purpose.
Use, copying, and distribution of any software described in this publication requires an applicable software license.
trademarks may be the property of their respective owners. Published in the USA [10/2/2017] [Best Practices Guide]
Dell believes the information in this document is accurate as of its publication date. The information is subject to change without notice.
3 Dell EMC PowerEdge and BlueData | version (2.0)
Table of contents Revisions............................................................................................................................................................................. 2
3.3 Data Storage ....................................................................................................................................................... 9
4.1 Local Storage .................................................................................................................................................... 11
5.7 Data Switches ................................................................................................................................................... 18
6.3 Data Storage and Platform Type ...................................................................................................................... 19
6.6 Isilon Model ....................................................................................................................................................... 20
A Tested Configuration Details ...................................................................................................................................... 22
4 Dell EMC PowerEdge and BlueData | version (2.0)
B Dell EMC Customer Solution Centers ........................................................................................................................ 24
C Technical support and resources ............................................................................................................................... 25
C.1 Related resources............................................................................................................................................. 25
5 Dell EMC PowerEdge and BlueData | version (2.0)
Executive Summary
The demand for businesses to make data-driven decisions has become no longer optional, but rather
required. The sheer number of choices in data processing platforms, both for structured and unstructured
data are overwhelming. Knowledge workers in customer’s lines of business are requiring IT organizations to
rapidly architect, provision, and maintain multiple sets of tools. Failure of an IT team to meet this demand
could see users trying to stand-up their own “unapproved / unmanaged” clusters or, at a greater risk, move
data out in un-governed public clouds where company intellectual property and customer data could be at
risk.
The benefit of partnering BlueData’s EPIC software with Dell EMC’s powerful portfolio of compute, storage,
and networking options allows our customers to consolidate multiple data analytics clusters into one shared
platform. This results in simplification of infrastructure, comprehensive control of data, and flexible and agile
capabilities around spinning up new and emerging technologies easily. This solution allows IT to focus on
helping their company differentiate themselves at the workload layer, rather than spending time at the
infrastructure plumbing layer.
6 Dell EMC PowerEdge and BlueData | version (2.0)
1 Introduction The demand for businesses to make data-driven decisions has become no longer optional, but rather required. The sheer number of choices in data processing platforms, both for structured and unstructured data are overwhelming. Knowledge workers in customer’s lines of business are requiring IT organizations to rapidly architect, provision, and maintain multiple sets of tools. Failure of an IT team to meet this demand could see users trying to stand-up their own “unapproved / unmanaged” clusters or, at a greater risk, move data out in un-governed public clouds where company intellectual property and customer data could be at risk.
Current Data Analytics Cluster State
The benefit of partnering BlueData’s EPIC software with Dell EMC’s powerful portfolio of compute, storage, and networking options allows our customers to consolidate multiple data analytics clusters into one shared platform. This results in simplification of infrastructure, comprehensive control of data, and flexible and agile capabilities around spinning up new and emerging technologies easily. This solution allows IT to focus on helping their company differentiate themselves at the workload layer, rather than spending time at the infrastructure plumbing layer.
7 Dell EMC PowerEdge and BlueData | version (2.0)
BlueData EPIC
8 Dell EMC PowerEdge and BlueData | version (2.0)
2 BlueData Epic Server Roles Below we describe the major server roles that are needed to stand up a production BlueData cluster. Updated sizing
guidance can always be found in the most recent documentation from BlueData.
2.1 Controllers These hosts are the initial installation point of BlueData and the key orchestrators in instantiating the containerized clusters on the Worker hosts. While a single host can act as the only controller (as well as a Worker host), we recommend that you deploy 3 Controller hosts as a master, standby, and arbiter highly available configuration. By default all controllers are provisioned as Workers as well, though some customer’s requirements may dictate that container-provisioning be disabled on the controller hosts in order to dedicate them to orchestration only. These hosts can also serve as network routers between the containerized internal overlay network and the outside world.
2.2 Workers Workers form the backbone of the compute infrastructure as well as provide the disk space on which the containerized hosts spin-up their scratch/ephemeral storage. In the case of the Local Storage configuration these Worker hosts will also hold the disks that will be dedicated to providing the Data Storage capacity. You’ll want a platform here that has multiple disk options of either large form-factor (LFF) disks that are suited to high-capacity or small form-factor (SFF) disks that emphasize performance. CPU cores and memory should be sized based off of the anticipated number of containers each host should be expected to handle in addition to 1 core and 1/8 allocated RAM +1GB for overhead. It’s important to understand that CPU cores can be oversubscribed, but memory can’t be.
2.3 Gateways Gateway hosts are optional components that provide the ability to surface network end-points for the internal containerized clusters. This allows designs to forgo the need for a routable container network through the controllers. Containerized clusters will terminate their user interface and service end-points on the front-facing IPs of the gateway hosts, mapped to ports above 10000. These hosts can be very light-weight configurations as well as even virtual machines on existing compute infrastructure. In the case of the Hadoop Ready Bundle solutions, you could deploy these on the existing Hadoop Edge Nodes.
3 Host Storage Types Each host in a BlueData cluster contains a couple of key types of local and/or remote storage. Some of these storage
types are optionally remote (Data Storage) and others are provisioned from local disk resources (Operating System and
Node Storage).
3.1 Operation System (OS) This is the storage on which the host-operating system is installed. All 3 roles (Controllers, Workers, and Gateways) have local Operating Systems that will need to be installed prior to BlueData package setup. We recommend, as with any production deployment, that you mirror the OS using RAID for resiliency. Check BlueData’s software requirements for the latest supported Linux OS revisions. Note that currently, the minimum amount of storage for the OS volume is 300GB, which can prevent the use of some OS/boot storage options like the Dell EMC Boot Optimized Storage Subsystem (BOSS) without manually overriding the installer.
3.2 Node Storage Node storage is the temporary/ephemeral/scratch space on each Worker host where the containers are instantiated. This is high-speed storage that is local to each Worker host. Regardless of deployment model, Node storage space is required. This must be un-formatted volumes that are presented to the Linux OS (e.g. /dev/sda, /dev/vda, /dev/nvme0n1). It’s important to remember that you must present each drive individually as opposed to in aggregate (RAID). In Dell PowerEdge RAID Controllers (PERC) this means flagging all the drives as “Non-RAID”. Detailed sizing considerations can be referred to in the BlueData System Architecture section of the documentation.
3.3 Data Storage The storage that each of the tenants use to store source data and results is referred to as Data storage. This can be made up of multiple local drives or of remote storage on an existing Hadoop (HDFS or Kerberos HDFS) and NAS (NFS). If using local storage, these drives must also be Non-RAID and unformatted. When using remote storage, you’ll use the BlueData DataTap mechanism to transparently attach over IP.
10 Dell EMC PowerEdge and BlueData | version (2.0)
Below is an example of how you can carve up the local drives on a Dell EMC PowerEdge R740XD to provide all 3 of the storage roles needed for BlueData. This just serves and one example of hardware configurations, further examples for additional use-cases and performance characteristics are detailed in Sections 4 and 5.
Example Disk Layout for Controller/Worker - Local Storage (High Performance)
GB
SA
S1
2G
b1
5k
GB
SA
S1
2G
b1
5k
GB
SA
TA
6G
bS
SD
GB
SA
TA
6G
bS
SD
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
TB
SA
TA
6G
b7
.2k
Driv
e 2
8
Driv
e 2
9
Driv
e 3
0
Driv
e 3
1
Driv
e 2
4
Driv
e 2
5
Driv
e 2
6
Driv
e 2
7
Driv
e 0
Driv
e 1
Driv
e 2
Driv
e 3
Driv
e 4
Driv
e 5
Driv
e 6
Driv
e 7
Driv
e 8
Driv
e 9
Driv
e 1
0
Driv
e 1
1
Driv
e 1
2
Driv
e 1
3
Driv
e 1
4
Driv
e15
Driv
e 1
6
Driv
e 1
7
Driv
e 1
8
Driv
e 1
9
Driv
e 2
0
Driv
e 2
1
Driv
e 2
2
Driv
e 2
3
Front Hot-Plug
Mid-Bay Flexbay
DataNodeOS
11 Dell EMC PowerEdge and BlueData | version (2.0)
4 Recommended Deployment Models The flexibility of the BlueData platform really shines when as you begin to look at the possible deployment models. Below
you’ll see 3 recommended models ranging from all local storage through to remote storage provided by either existing
Hadoop clusters or Dell EMC Isilon’s scale-out NAS capabilities. Which model you choose should be determined after
consultation with your Dell EMC and BlueData sales support resources (as well as the Dell EMC Customer Solutions
Center)
4.1 Local Storage In this model, each Worker and Controller/Worker has local storage for all three storage types: OS, Node, and Data resulting in a more simplified architecture. The key differentiator in this deployment model is the local storage is allocated as Tenant data. Local disks (usually rotational) are used to make up the BlueData HDFS instance that is shared across all Tenants. Each tenant is restricted from viewing the other tenant’s files, but they are all resident on the single HDFS namespace that BlueData provides. Different server models can be used here to provide a platform that is optimized for either high-capacity (LFF) or high performance (SFF). However, we do recommend that all Worker hosts have the same disk configuration and layout.
Dell EMC PowerEdge R740XD with Large Form-factor Drives (Top) and Small Form-factor Drives (Bottom) – Bezels Removed
12 Dell EMC PowerEdge and BlueData | version (2.0)
Local Storage Deployment Model
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
Sta
ck ID
Sta
ck ID
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
OS Node Data
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
Controllers / Workers
Workers
(Master, Standby, Arbiter)
Data Switches
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
Management Switches
iDR
AC
s
Gateway(s)(Optional)
CorporateNetwork
13 Dell EMC PowerEdge and BlueData | version (2.0)
4.2 Remote Storage on Dell EMC Hadoop Ready Bundle Many customers have existing, robust, hardened primary Hadoop clusters that they would like to leverage for storage. In this deployment model you can use BlueData as the Controller/Compute tier and leave your source data and result data on the remote HDFS solution. Dell EMC has a long history of providing validated reference configurations of both Cloudera and Hortonworks distributions of Hadoop. These solutions can save IT departments hours of valuable time by taking care of all the steps of racking, provisioning the operating systems, and installing and configuration the Hadoop distribution, allowing IT to spend time focused on the workloads that differentiate their company.
Remote Storage on Dell EMC Ready Bundle Deployment Model
Sta
ck ID
Sta
ck ID
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
OS Node
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
Workers
(Master, Standby, Arbiter)
Data Switches
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
Management Switches
DataStorage
Name Node(s)
iDR
AC
s
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
Gateway(s)(Optional)
Controllers / Workers
CorporateNetwork
14 Dell EMC PowerEdge and BlueData | version (2.0)
4.3 Remote Storage on Dell EMC Isilon Dell EMC Isilon’s multi-protocol scale-out storage capabilities can be leveraged to provide a remote filesystem with a significantly lower total cost of ownership (TCO). By reducing the need for multiple software replicas and sharing the same data across SMB/NFS/HDFS Isilon can reduce the complexity of a remote storage solution while allowing the introduction of a robust portfolio of data management options.
Remote Storage on Dell EMC Isilon Deployment Model
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
Sta
ck ID
Sta
ck ID
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
OS Node
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
TBSATA
6Gb7.2k
Workers
(Master, Standby, Arbiter)
Data Switches
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
Management Switches
DataStorage
iDRA
Cs
Gateway(s)(Optional)
Controllers / Workers
CorporateNetwork
15 Dell EMC PowerEdge and BlueData | version (2.0)
5 Component Builds
5.1 Recommended Host Counts
Components Local Storage Remote Storage on Dell EMC Hadoop
Ready Bundle
Remote Storage on Dell EMC Isilon
Controller/Worker Hosts (with local data storage)
3 (High Capacity or Performance)
N/A N/A
Worker Only Hosts (with local data storage)
7 (High Capacity or Performance)
N/A N/A
Controller/Worker Hosts (without local data storage)
N/A 3 3
Worker Only Hosts (without local data storage)
N/A 7 7
Gateways 2 2 2
Management Switches 2 2 2
Data Switches 2 2 2
External Storage Solution N/A Dell EMC Ready Bundle for Hadoop
5.5 Gateways (Optional) The optional gateway hosts can be a very light-weight server whose configuration is highly dependent on the customer use-case. A non-specific configuration is listed below that details the recommended CPU core count and RAM amounts.
Component
CPU Core Count 8-Cores
Memory 32GB
OS Storage 100 GB
Gateways (Optional)
GB
SA
S1
2G
b1
5k
GB
SA
S1
2G
b1
5k
GB
SA
TA
6G
bS
SD
GB
SA
TA
6G
bS
SD
18 Dell EMC PowerEdge and BlueData | version (2.0)
5.6 Management Switches
Component
Switch Model Dell EMC Networking S3148-ON
Management Switches
5.7 Data Switches
Component
Switch Model Dell EMC Networking S4148F-ON
Data Switches
27 29 31 33 35
28 30 32 34 36
37 39 41 43 45 47
38 40 42 44 46 48
LNK ACT7 9 11
2 4 6 8 10 12
13 15 17 19 21 23
14 16 18 20 22 24
25
26
Stack No.
1
2
49 50SFP+
LNK ACT
5
COMBO PORTS 47 48
Sta
ck ID
19 Dell EMC PowerEdge and BlueData | version (2.0)
6 Additional Configuration Considerations
6.1 Host Count This document outlines 3 configurations that allow for 3 different methods for storing persistent data. The amount of compute resources in each configuration have been kept the same (processor core count and RAM) so that each configuration can instantiate the same number of containerized applications. However, it should be noted that in the remote storage configurations (Remote Hadoop and Dell EMC Isilon), since the local storage processing overhead of offloaded to those storage systems, the amount of compute and memory needed could diminish. With that in mind, you could see the same workload performance in those configurations as with the local storage configuration, all while allowing for a lower total host count.
6.2 Node Storage The drives providing Node storage are the best place to make an investment in speed. This can range from faster / larger SSDs on through PCI-E NVMe devices with microsecond latency. Improving the speed here will improve the speed of provisioning containers, the speed at which those containers can execute storage operations, and reduce the latency of transactions. The sizing of this space will need to take into account how many containers of each solution stack will need to be deployed on each host. The advantage of using containerization here is that you’re not penalized for the OS portion of each host since they are shared amongst all other hosts that use that same container’s OS. This means that the only thing that distinguishes each container from each other are the deltas to the base image.
6.3 Data Storage and Platform Type The amount of necessary data storage will vary depending on customer use-case. The Dell EMC PowerEdge R740XD compute platform allows for multiple configurations across both the performance and capacity dimensions. However, there are a number of additional optional platforms you could consider: Controller/Worker –Local Storage
> Cost-Optimized Local Storage–Dell EMC PowerEdge R540
> Capacity-Optimized Local Storage – Dell EMC R740XD with larger drives (limit to less than
100GB raw per host as a best practice)
> Density-Optimized Local Storage– Dell EMC PowerEdge R640
> Modular Local Storage–Dell EMC PowerEdge FX2 with Dell EMC PowerEdge FC640 and
> Modular Compute Only –Dell EMC PowerEdge FX2 with Dell EMC PowerEdge FC640
> Hyper-scale – Dell EMC PowerEdge C6420
6.4 Host Networking The amount of bandwidth delivered to each host should take into consideration both storage traffic (host to host if using local storage and remote storage traffic if using a remote solution like an existing Hadoop or Isilon). In our recommended configurations, we allocated a bonded set of four (4) 10GbE links for both throughput and failover accommodations. There are many additional ways to deliver substantial bandwidth to Dell EMC PowerEdge platforms including multiple 40GbE
20 Dell EMC PowerEdge and BlueData | version (2.0)
links as well as the emerging 25GbE solutions. 25GbE shows particular promise as it’s cost:performance ratio is unmatched in the other technologies. As you scale this solution to multiple racks, we recommend using a leaf/spine design to allow for scale-out capabilities. You can look to Dell EMC’s wide portfolio of aggregation tier switches at 40GbE and 100GbE to provide the spine for your topology. In our recommended configurations, each host is delivering both the Container and the Management traffic on the same interface. You could split this using VLAN tagging or even divide the 4x 10GbE ports into 2x 2x10GbE trunks with one trunk handling management, and one trunk handling the container traffic.
6.5 Ready Solution Speed / Capacity Dell EMC’s Ready Solutions for Hadoop offer multiple scaling options for performance and modularity. When implementing this solutions you have autonomy to scale the Ready Solution storage independent from the BlueData compute solution. If you are looking to improve remote storage speed, you could either scale the network or Hadoop worker host disk speed. If you are looking for capacity, you can scale the number of Hadoop worker hosts in the remote solution or increasing the drive capacity in the worker hosts.
6.6 Isilon Model With a broad portfolio of models, the Dell EMC Isilon product family has many options to help you optimize your solution for speed or capacity as well as any mix of the two. All flash models can significantly improve storage latency. Hosts with larger, slower drives can help optimize your solution for storage capacity without the need for the physical space that 3 replicas requires in the Local Storage configuration.
> Capacity Focused – Dell EMC Isilon H500 each with 15x 4TB SATA and 2x 1.6TB SSD
> Performance (Disk) Focused – Dell EMC Isilon H600 each with 30x 1.2TB SAS and 2x
1.6TB SSD
> Performance (All-Flash) Focused – Dell EMC Isilon F800 each with 15x 3.2TB SSD
21 Dell EMC PowerEdge and BlueData | version (2.0)
7 Conclusion As many organizations are finding out, flexibility and agility are keys to success when it comes to data analytics. This recommended configuration for BlueData on Dell EMC PowerEdge Servers, Ready Solutions, and Isilon provides the ability to instantiate many of the latest and greatest data analytics tools of today, while providing the infrastructure to build on for the tools of tomorrow. Consolidating your data analytics resources into a single shared resource pools provides the ability to keep data under control and resources efficiently utilized, which then allows you to focus your resources where it matters the most, not on the infrastructure plumbing.
22 Dell EMC PowerEdge and BlueData | version (2.0)
A Tested Configuration Details
The below configuration was stood up in the Dell EMC Customer Solution Centers to perform basic validation
routines. This testing consisted of running in both Local Storage mode as well as attaching to an existing
HDFS filesystem provided by a Dell EMC Ready Solution for Cloudera Hadoop cluster.
Component
Server Model Dell EMC PowerEdge R740XD
Chassis Chassis with up to 24x2.5" HDD including 12 NVMe Drive Bays (not used)