jetnexus ALB-X 2.0 Performance Test A Broadband-Testing Report
First published May 2012 (V1.0) Published by Broadband-Testing Tel : +376 633010 E-mail : info@broadband-testing.co.uk Internet : HTTP://www.broadband-testing.co.uk 2012 Broadband-Testing All rights reserved. No part of this publication may be reproduced, photocopied, stored on a retrieval system, or transmitted without the express written consent of the authors. Please note that access to or use of this Report is conditioned on the following: 1. The information in this Report is subject to change by Broadband-Testing without notice. 2. The information in this Report, at publication date, is believed by Broadband-Testing to be accurate and reliable, but is not guaranteed. All use of and reliance on this Report are at your sole risk. Broadband-Testing is not liable or responsible for any damages, losses or expenses arising from any error or omission in this Report. 3. NO WARRANTIES, EXPRESS OR IMPLIED ARE GIVEN BY Broadband-Testing. ALL IMPLIED WARRANTIES, INCLUDING IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT ARE DISCLAIMED AND EXCLUDED BY Broadband-Testing. IN NO EVENT SHALL Broadband-Testing BE LIABLE FOR ANY CONSEQUENTIAL, INCIDENTAL OR INDIRECT DAMAGES, OR FOR ANY LOSS OF PROFIT, REVENUE, DATA, COMPUTER PROGRAMS, OR OTHER ASSETS, EVEN IF ADVISED OF THE POSSIBILITY THEREOF. 4. This Report does not constitute an endorsement, recommendation or guarantee of any of the products (hardware or software) tested or the hardware and software used in testing the products. The testing does not guarantee that there are no errors or defects in the products, or that the products will meet your expectations, requirements, needs or specifications, or that they will operate without interruption. 5. This Report does not imply any endorsement, sponsorship, affiliation or verification by or with any companies mentioned in this report. 6. All trademarks, service marks, and trade names used in this Report are the trademarks, service marks, and trade names of their respective owners, and no endorsement of, sponsorship of, affiliation with, or involvement in, any of the testing, this Report or Broadband-Testing is implied, nor should it be inferred. ii Broadband-Testing 1995-2012
TABLE OF CONTENTS TABLE OF CONTENTS... III BROADBAND-TESTING... V EXECUTIVE SUMMARY... 1 INTRODUCTION: PERFORMANCE AND FLEXIBILITY... 2 ALB-X 2.0 - KEY FEATURES... 3 flightpath... 4 ALB-X: PUT TO THE TEST... 9 The Test Bed... 9 Test 1: Basic, One-Armed Layer 7 Throughput... 10 Test 2: Cached, Basic Layer 7 Throughput... 10 Test 3: Uncached Layer 7 Transactions Per Second Capability... 11 Test 4: Cached Transactions Per Second Capability... 12 Test 5: flightpath Tests... 13 SUMMARY & CONCLUSIONS... 15 TABLE OF FIGURES Figure 1 ALB-X 2.0 Appliance...3 Figure 2 ALB-X 2.0 GUI Home Page...3 Figure 3 flightpath Condition Example...4 Figure 4 Drag n Drop flightpath...5 Figure 5 ALB-X 2.0 Server Monitoring...6 Figure 6 Configuring Acceleration...6 Figure 7 Configuring Compression and Cache...7 Figure 8 ALB-X 2.0 Reporting Stats...7 Figure 9 ALB-X 2.0 Dashboard...8 Figure 10 Spirent Testcenter...9 Figure 11 One-Armed Layer 7 Throughput... 10 Figure 12 Cached Layer 7 Throughput... 10 Figure 13 Cached Test: Network Response Time... 11 Figure 14 Uncached TPS Performance... 11 Figure 15 Cached TPS Performance... 12 Figure 16 Cached TPS: Network Response Time... 12 Figure 17 flightpath Rule In Place: No Caching... 13 Broadband-Testing 1995-2012 iii
Figure 18 flightpath Rule In Place: Caching Enabled... 13 Figure 19 flightpath With Caching: Network Response Time... 14 Figure 20 flightpath With Caching: URL Response Time... 14 iv Broadband-Testing 1995-2012
BROADBAND-TESTING Broadband-Testing is Europe s foremost independent network testing facility and consultancy organisation for broadband and network infrastructure products. Based in Andorra, Broadband-Testing provides extensive test demo facilities. From this base, Broadband-Testing provides a range of specialist IT, networking and development services to vendors and end-user organisations throughout Europe, SEAP and the United States. Broadband-Testing is an associate of the following: NSS Labs (specialising in security product testing) Limbo Creatives (bespoke software development) Broadband-Testing Laboratories are available to vendors and end-users for fully independent testing of networking, communications and security hardware and software. Broadband-Testing Laboratories operates an Approval scheme which enables products to be short-listed for purchase by end-users, based on their successful approval. Output from the labs, including detailed research reports, articles and white papers on the latest network-related technologies, are made available free of charge on our web site at HTTP://www.broadband-testing.co.uk Broadband-Testing Consultancy Services offers a range of network consultancy services including network design, strategy planning, Internet connectivity and product development assistance. Broadband-Testing 1995-2012 v
EXECUTIVE SUMMARY Application Delivery Control (ADC) and application acceleration solutions now need to be more flexible than ever, thanks to the influence of the cloud, SaaSbased application delivery and changing data centre profiles. With the 2.0 version of jetnexus ALB-X load-balancer/adc product, we have a true range of solutions that provides this level of flexibility in formats to satisfy every need large or small. While the primary focus of vendors in this market to date has been high-end, meaning the vast majority of companies in the small to medium business (SMB) sector (93% of companies globally) have been unable to afford the benefits of load-balancers, with both the virtual appliance and the ALB-X ISO version, finally we have a true mass market product and one that comes with green credentials. Regardless of which version of the product is used, it is simplicity itself to deploy, keeping costs down in the long as well as short term. This low cost of ownership, both in terms of CapEx and OpEx, means that the ALB-X 2.0 has real credentials from a hosted/managed service perspective for service providers as well as for direct enterprise and SMB use. The technology is applicable to any user with high value data and the need to maintain uptime. In addition to looking to improve performance notably in https environments with 2.0, jetnexus has looked to add far more flexibility than before, both in terms of its routing/interface support capabilities and with the introduction of flightpath, a powerful routing engine which enables clients to create and implement bespoke application rules. The aim here is to considerably improve day-to-day traffic management for clients, giving them greater control and intelligence in managing service delivery. The management GUI has also been completely refreshed to make the product easier to manage while integrating the new features. Performance wise, we put a mid-range ALB-X 2.0 appliance to the test using 10GbE interfaces and tested in a variety of scenarios with real-world http traffic. We achieved bi-directional linerate at 10Gbps without caching and linerate with caching enabled while carrying out true Layer 7 content changing operations. We also saw in excess of 57,000 transactions per second achieved in cached mode and around 35,000 transactions per second in un-cached mode. As we enter the next generation of the Internet, the reliance on access to data held anywhere across the globe becomes more critical than ever, as does the need for affordable, easily managed ADC and application acceleration technology. With 2.0 of ALB-X, in all its forms, jetnexus is looking to provide exactly that. Broadband-Testing 1995-2012 1
INTRODUCTION: PERFORMANCE AND FLEXIBILITY Load-balancing (L-B) and application delivery control - ADC - is all about optimising data access and availability. At the same time, it is not - in 2012 - about excessively expensive and overly-complex optimisation appliances. Server prices are now so competitive, that the days of six-figure sums for buying load-balancing appliance are surely over for all but the largest data centres? So, striking a price:performance:features balance is far from trivial to achieve. Meantime, the data itself is changing. Everyone is talking about "big data" but the broader picture is simply that, both the data itself and the way it is being access (and where it is accessed from) are changing and becoming more varied, so flexibility in an L-B/ADC device is a more critical requirement than ever before. Sounds like a big ask then - competitive pricing, more than sufficient performance, high levels of reliability/redundancy and high levels of data control - all in one appliance, whether physical or virtual. Such is the challenge that UK-based jetnexus has accepted and whose ALB-X 2.0 appliance we put to the test here. Moving The Game On So Everyone Can Play... Historically, the benefits of load-balancing and application acceleration, while clearly advantageous to any user who accesses data and applications stored in a data centre or server farm and who doesn t have been restricted to Enterprise and above. However, with its ALB-X product, jetnexus made these features available in every format from a physical appliance through virtual appliance, down to ISO an image sitting on a CD or USB key that turns pretty well any server into a load-balancer. Now the company has introduced version 2.0 of its product, with the additional features available in all three product formats. Regardless of the format, all ALB-X 2.0 products come with very competitive price tags compared with traditional L-B solution providers. The cost of server real estate has fallen so far since the heady days of the dot com boom, it is not reasonable now to demand mega-money for an ADC/L-B solution if it's simply cheaper to add more servers and other forms of redundancy. With a new GUI design for 2.0, the company is also making it easier to manage this environment than before, while adding new levels of flexibility and performance (look out for our follow-up performance test reports on the ALB-X 2.0). What jetnexus is doing, then, is taking both the excess cost and complexity out of ADC and application acceleration, while at the same time moving the game on. Whether you re a tin man and opt for the turnkey appliance, a virtualisation fan and go for the virtual appliance, or simply want to turn a surplus server into an ADC you have the choice. Note - with the ALB-X VA, the virtual appliance version of the ALB-X product, this offers the full range of ALB-X features, just in a virtual appliance. The build includes a custom kernel operating system integrated with the ALB-X software. The virtual appliance runs on many desktop environments as well as the increasingly ubiquitous VMware ESX server. 2 Broadband-Testing 1995-2012
From a user perspective, then, as a small to medium business (SMB), the benefits of L- B/ADC are truly affordable. At the opposite end of the scale, if you re a Service Provider or Hosting company looking to offer cloud-based or managed services to a broad range of customer types and sizes then, again, all the options are available. Equally, if you re an enterprise with HQ and a number of satellite or branch offices, there is a complete solution available to meet these diverse needs. Figure 1 ALB-X 2.0 Appliance So let us now look at the features and functionality of the ALB-X 2.0, highlighting the key new features within. ALB-X 2.0 - KEY FEATURES Here we focus on what we see as the key features being offered by jetnexus ALB-X 2.0 that are of real use by a broad range of customer types to provide Load-Balancing and Application Delivery Control. New to 2.0 of the product, in addition to significant performance improvements on the hardware side, a new GUI replaces the old version with a far sharper appearance than previously. Figure 2 ALB-X 2.0 GUI Home Page Broadband-Testing 1995-2012 3
Routing options are also improved significantly so that multiple interfaces and combination can now be used to provide both more flexibility and better performance than before. flightpath An important addition is flightpath - a powerful routing engine which enables clients to create and implement bespoke application rules. The aim here is to considerably improve day-to-day traffic management for clients, giving them greater control and intelligence in managing service delivery. This is especially crucial when managing online applications and ecommerce sites. Due to the configurable nature of flightpath, rule options are infinite but some common uses are to, for example, block unwanted requests, rewrite user requests, control spiders and fix broken URLs. FlightPATH can be used to modify incoming and outgoing HTTP/S content and requests. As well as using simple string matches such as starts with and ends with, powerful Perl compatible regular expressions can also be implemented for more control. In addition, custom variables can be created and used in the action enabling many different. Figure 3 flightpath Condition Example Some common examples of real-world applications of flightpath are suggested below though please note, these are purely a handful of options in reality the possibilities are all but infinite: For application firewalling and security Block unwanted IP s Force user to HTTPS for specific (or all) content Block or redirect spiders Prevent and alert cross site scripting Prevent and alert SQL injection Hide internal directory structure Rewrite cookies Secure directory for particular users For application management Redirect users based on path Provide Single sign on across multiple system Segment users based on User ID or Cookie Add headers for SSL offload Language detection Rewrite user request Fix broken URL s Prevent directory access/ browsing Send spiders different content 4 Broadband-Testing 1995-2012
An important consideration with flightpath is that it doesn t require development skills. You create rules using Action and Condition options and build your rule in a similar way that you would create a search rule in a reporting environment. This means that administrative, as well as IT, individuals are suited to development flightpath rules, making the product extremely flexible. Simple drag n drop techniques can be used to select variables and build rules accordingly. Figure 4 Drag n Drop flightpath In terms of what else ALB-X 2.0 offers, let s start with the obvious. Performance is clearly a primary requirement for L-B and ADC. For this reason, ALB-X 2.0 includes Content Caching a feature that has typically only been found on dedicated devices or top end application delivery controllers, and then typically as a high-cost option. Is it effective? Well, one e-commerce customer of jetnexus has reported that it now serves 94% of all its content from the cache. Not only does this improve performance but also reduces the number of back end servers and subsequent licences required. Layer 7 L-B is fundamental to ADC, operating, as it does, at the application protocol level. Again, this was a feature absent on lower-priced devices until the likes of jetnexus introduced at an entry-mid level. Key to Layer 7 L-B is that the device recognises the type of data it is processing - HTTP and HTTPS where the ALB-X 2.0 is concerned and consequently allows for many features that are not possible at lower-level L-B such as URL switching, cookie-based persistence and request scanning. In conjunction with Layer 7 optimisation, the ALB-X 2.0 also supports Connection Management. This is used to optimise the performance of TCP/IP in that it terminates the thousands of TCP connections applications generate at the device, then consolidates them to send far less (effectively bundled) connection requests to the web server. TCP connections are one of the primary killers of server performance placing, as they do, enormous stress levels on the server pool. Connection Management is therefore a fundamental requirement of high performance L-B and application delivery. Broadband-Testing 1995-2012 5
How do you know if all servers in the pool are operating in full health? Unless you monitor the server pool, it is impossible to know, meaning you could be forwarding large amounts of traffic to a server that is experiencing operational problems, or is simply saturated. For this reason, the ALB-X performs a series of server health checks, from simple ping and TCP tests to full web GETs, to ensure every server is performing as it should. Figure 5 ALB-X 2.0 Server Monitoring Another fundamental of data performance optimisation is accelerating web traffic. The ALB-X 2.0 uses jetnexus compression algorithm to deliver compressed, optimised content in a form that is readable by the browsers. This acceleration is tuneable via the ALB-X 2.0 management GUI and rule of thumb guidelines are provided by jetnexus to aid with optimising configuration for each particular environment. Figure 6 Configuring Acceleration Rule based compression detects and compensates for browser behavioural differences and defects, therefore guaranteeing that accelerated pages can be displayed correctly by the client s browser. It has the ability to accelerate most content including Javascript, AJAZ, XML and other types, as well as Office documents (e.g. DOC, XLS, PPT, etc.) and configurable MIME types. 6 Broadband-Testing 1995-2012
Figure 7 Configuring Compression and Cache It can act on HTTP headers - POSTS, SOAP, HTML, JavaScript, etc and has the ability to accelerate HTTP 1.1 through HTTP 1.0 Proxies and Caches (e.g. Squid, MS ISA, etc.) and can accelerate through Personal Firewalls (e.g. Symantec, MacAfee, etc.). Streaming compression is also supported. In conjunction with connection control, terminating SSL (https) requests at the Load- Balancer, rather than at the server SSL offload in other words massively reduces the performance impact on the server pool. Past testing has shown that https traffic can generate 10 times more load on the server than regular http traffic. Therefore, by terminating it at the ALB-X device, not only is server performance dramatically improved but it also means that in some cases server consolidation can take place, reducing costs further. This also has major management benefits both from an operational and cost perspective. Figure 8 ALB-X 2.0 Reporting Stats Broadband-Testing 1995-2012 7
A number of reporting and logging features are available with the ALB-X 2.0. Reporting is where many network devices typically come unstuck if it is impossible to get decent statistics out of the device then how do you know if it performing as well as it should be? For example, you can capture live traffic stats to see what levels of content compression, throughput rates and cache performance and view it real-time. Equally, you can monitor the status of the ALB-X 2.0 device itself. As well as being available in tabular format, a dashboard can be viewed, providing a graphical analysis of current device status, with alert logs and compression performance. Figure 9 ALB-X 2.0 Dashboard In addition to the aforementioned logging, statistics and reporting on inbound and outbound traffic in real-time, jetnexus provides extensive alerting facilities for example, via an email alert as well as full SNMP support. 8 Broadband-Testing 1995-2012
ALB-X: PUT TO THE TEST The Test Bed To put the ALB-X 2.0 to the test, we created a test bed using Spirent's Testcenter, to generate web traffic with both IPv4 and IPv6. Figure 10 Spirent Testcenter This was loaded with a Spirent 10Gigabit module, featuring 2 x 10Gigabit fibre FXP+ ports. To this we directly connected the ALB-X 2.0 ports - one client side and one server side. Our ALB-X 2.0 was a Model E4, based on a quad core architecture with 4GB RAM, four gigabit and two 10Gigabit fibre modules (these are interchangeable with alternative connector types) and a 70GB SSD. Our aim was to run real-world HTTP traffic, along with.gif graphics for the cache-based tests, in order to test both throughput and transactions per second (TPS) capabilities, both key metrics with this kind of product. Our test network was configured with two networks each on the client and server side, with both client and server pools defined for the Spirent equipment to emulate. Note, these are real, pingable HTTP clients and servers, passing real traffic. On the ALB-X 2.0 we created a configuration with a simple Layer 7 rule to use during the basic testing and a more complex rule in order to test performance with flightpath - jetnexus' scripting tool - in action. Throughout the testing we did not adjust the basic device configuration, so this was a real-world test scenario, in line with how an IT department would deploy the ALB-X, rather than reconfiguring optimally for each test as is sometimes the case; hardly real-world in that scenario... For our throughput-focused tests we created http traffic using a 100KB transaction file. This would naturally limit the number of transactions per second achievable compared with using a much smaller file size, but did enable us to reach maximum potential throughput as quickly as possible, while still using a realistic file size. In all cases we were looking to push the device under test until we saw packet loss occur, so results are based on zero packet loss. For the transaction-oriented tests we created http traffic using a 64-byte packet size, as is the industry norm for these kinds of tests while, for the caching tests we created a 250KB GIF file. Broadband-Testing 1995-2012 9
Test 1: Basic, One-Armed Layer 7 Throughput The way the ALB-X is designed, in non-cached mode for our testing, we were measuring performance in a one-armed mode effectively since, with our test configuration, the ALB- X has to pass the traffic twice, effectively meaning bandwidth is split between the two actions. So, with a 10Gbps connection we were looking to achieve as close to 5Gbps as possible on each measurement point. Figure 11 One-Armed Layer 7 Throughput We actually achieved 4.5Gbps consistently, meaning a 9Gbps+ total throughput capability - close as is realistic to line rate at 10Gbps, allowing for the fundamental overheads IP and Ethernet generate. Zero packet loss was encountered and the management stats on the ALB-X showed significant CPU overhead was still available for other operations - a common theme throughout the tests. Test 2: Cached, Basic Layer 7 Throughput Figure 12 Cached Layer 7 Throughput 10 Broadband-Testing 1995-2012
In cached mode, with our GIF files, we were able to serve directly from the cache and achieve a consistent throughput of around 9.13Gbps - again as close as is realistically possible to line rate. Again, we had zero packet loss and significant CPU overhead available still at the ALB-X. Figure 13 Cached Test: Network Response Time Given the primary roles of L-B/ADC devices - often sat in front of web and application servers - the ability to deliver pages and files as quickly as possible is critical. Throughout this test we saw excellent response times - a round trip time of just 0.88ms on average for the server response (from cache in this case) meaning instantaneous file delivery. Test 3: Uncached Layer 7 Transactions Per Second Capability Figure 14 Uncached TPS Performance Using a 64byte file with caching disabled, we were able to regularly see just under 35,000 TPS achieved, actually peaking at 34,779 TPS. In this case we did see some packet loss - 0.79% - but this is primarily a factor of the test configuration; allowing longer would Broadband-Testing 1995-2012 11
enable the TCP connections to all successfully close (see graph) so we would not expect to encounter any lost packets in the real-world at these performance levels, as spare CPU capacity on the ALB-X, again, indicated. Test 4: Cached Transactions Per Second Capability Figure 15 Cached TPS Performance With caching enabled, we were able to sustain around 50,000 TPS and peak at 57,333 TPS, with zero dropped packets and generating up to 5,568 TCP connections per second. Figure 16 Cached TPS: Network Response Time Again, network response times were excellent despite the number of transactions being carried out, with an average round trip time of 3.66ms and a maximum of 9.74ms. 12 Broadband-Testing 1995-2012
Test 5: flightpath Tests In order to further test the Layer 7 performance capabilities of the ALB-X we added a flightpath rule that carried out a full body content replacement - reading in, identifying the content, then modifying it - an extensive Layer 7 operation. Figure 17 flightpath Rule In Place: No Caching Testing first with caching not in place we were able to achieve a maximum of 19,215 TPS (effectively complete page rewrites) with a very small percentage of lost packets, again due to the test not being long enough to enable all TCP connections to close correctly, with spare CPU capacity on the ALB-X. Enabling caching saw our peak TPS figure increase dramatically to 45,328 TPS with zero unsuccessful packets and a peak TCP connections per second rate of 4,370. Bearing in mind that these are complete page rewrite operations, this is an exceptional result that shows the true benefits of caching when it is correctly enabled. Figure 18 flightpath Rule In Place: Caching Enabled From a "user experience" perspective, more to the point are the excellent network and URL page response times recorded. Broadband-Testing 1995-2012 13
Figure 19 flightpath With Caching: Network Response Time An average round trip time of just 4.47ms is excellent, with a maximum of 13.27ms and a minimum of just 0.21ms. Figure 20 flightpath With Caching: URL Response Time URL response times were equally good - 5.75ms on average was the time it took to deliver a completely re-written page from cache. In the real world this is effectively instantaneous. Overall, performance was consistently good, whether with or without cache enabled. In terms of dropped packets, the ALB-Xs resistance to dropping packets throughout the testing was excellent, implying high levels of reliability in real deployments. 14 Broadband-Testing 1995-2012
SUMMARY & CONCLUSIONS Application Delivery Control (ADC) and application acceleration solutions now need to be more flexible than ever, thanks to the influence of the cloud, SaaS-based application delivery and changing data centre profiles. With its original ALB-X product, jetnexus already offered excellent flexibility for L-B and ADC deployment, notably in terms of the form factors available physical, virtual and ISO. Now, with the release of version 2.0 and the addition of flightpath, the appliance in any format can be requested to carry out any number of actions and via a simple drag n drop interface that removes the requirement for programming skills. The GUI has also been improved significantly so all features are more readily accessible, while allowing the new features to be integrated. Performance wise, we succeeded in achieving effective Layer 7 line rate at 10Gbps, while - in terms of transactions per second measurements - both recorded TPS rates in uncached and cached modes and response times (network and URL/page) were above expectations. In terms of dropped packets, the ALB-Xs resistance to dropping packets throughout the testing was excellent - well above normal, so we would anticipate seeing high levels of reliability in real world deployments with this appliance. Finally, from a price:performance level, especially when allowing for new features such as the flightpath capabilities, with the ALB-X 2.0 jetnexus is offering a huge amount for a low outlay. As such we recommend you take a look at it. Broadband-Testing 1995-2012 15