Wednesday, December 17, 2014

Search - Splunk, Elasticsearch , Log Management - Sumologic , logstash

Elasticsearch says it's on a mission to make massive amounts of data available to businesses. Toward that end, in February the company debuted release 1.0 of its Elasticsearch search and analytics software for working with big data. The software is part of the company's open-source ELK stack that includes Elasticsearch, Logstash (log management) and Kibana (log visualization) software.
Elasticsearch is competing with Solr, the open-source search technology from the Apache Software Foundation. Many reviewers, users and solution providers give both technologies high marks. So it will be interesting to see if both gain traction, or if one or the other wins out.


  • Logstash is an agent to collect and mine logs : https://github.com/logstash-plugins
  • Also see sumologic where they claim they have their  log management agent in AWS



Risk Focus is an expert implementation partner of Splunk in the financial services arena and a big proponent of the value Splunk delivers.  We have implemented Splunk in numerous banks, funds and clearing/exchange firms across multiple mission-critical use cases.
Yet Splunk is not the only log analyzer out there.  In particular, ELK (elasticsearch + logstash + kibana) is a growing open source offering which purports to have the same functionality as Splunk at virtually no cost.
Is this a valid claim? Do Splunk and ELK provide similar functionality, stability and technical richness needed by corporate institutions like banks, asset managers, hedge funds, exchanges, industry utilities and major technology providers?

Cost of Splunk vs. ELK

Let’s start by taking a look at the overall cost of investing in Splunk versus ELK.  Corporate buyers invariably look at total cost of ownership (TCO) when making their buying decision because they want to know the true “all in cost” not just the up front fee.
“Splunk is expensive and ELK is free.”
A web search will turn up hundreds of blog entries claiming Splunk’s pay-per-gigabyte-indexed pricing model is expensive. Splunk data indexing charges sound pricey, but the way the pricing actually works is far cheaper than it first appears. Yes, an ELK license is free but Splunk is amazingly cheap, too.
Underlying the “Splunk is expensive” claim is the assumption that all data will be indexed, which is rarely true.  A proper implementation includes an up-front analysis such that only the valuable subset of a company’s data is indexed.
Midsize and larger companies tend to purchase software and data licenses at bulk discounted rates.  This gives a discount off list price and provides predictability (no “bait and switch” surprises) after adoption. For less than the cost of a single skilled FTE in a G10 country you can index a huge amount of log data with Splunk across your IT infrastructure and earn tremendous efficiency cost savings. We’ve seen Splunk’s rates dropping over time, so it’s getting even cheaper. If you just need to dabble, do basic development and testing, or a proof of concept, Splunk offers a free Enterprise license up to 500MB per day.
The primary concern for sophisticated corporate buyers is cost-to-value  or total cost of ownership (TCO).  Data license costs (at least at the pricing level of Splunk or other log analyzers) tend to be the least important factor in the TCO equation.  The value of cost savings and new revenue discoveries can provide immense financial value that dwarf the license costs.
“Once we get ELK going it will be cheaper. The time to learn and get ELK figured out is an investment.”
Both Splunk and ELK can be installed and running quickly at a basic level with minimal learning time. Efficiency gains from reducing “Team GREP” activities (which never really identify or resolve underlying issues well), the actionable intelligence obtained (allowing you to optimize IT spend very quickly), and operational risk reductions all argue in favor of installing a log analyzer as soon as possible.
Once installed, the question then becomes “can Splunk or ELK knowledge can be rolled out through an organization quickly?”  Rapidly enabling multiple users to take advantage of indexing, correlation and dashboarding capabilities is necessary to generate business value from a log analyzer.
Splunk is far ahead of ELK in speed of roll out and depth of coverage.  Splunk offers a rich education program, a Professional Services group and an expansive network of skilled consulting partners.  Getting a team Splunk certified takes less than 1 month.  Hiring a Splunk partner firm to roll out capabilities quickly and build advanced correlation apps can further shortcut the time-to-value.  Splunk has a large App Store with hundreds of free and paid apps to connect to standard IT hardware and software platforms. ELK is growing rapidly and is making similar education efforts, but is years behind Splunk in these critical areas.
“Beware the hidden costs.”
Compared to Splunk, we believe the investment required to institutionalize ELK is far more time-consuming and costly in terms of lost efficiencies and investment dollars than buying Splunk.  An ELK user is likely to find that they must create an entire infrastructure around knowledge transfer, skill-building and connectors to underlying log sources before rolling ELK out across the firm.These are hidden but serious costs of choosing ELK.  The incremental cost of a Splunk data license is much lower then the time and costs of building an in-house knowledge and support structure from scratch around ELK.
For a small organization with basic needs and a small IT support group, ELK is certainly a good investment.  It’s free from a license standpoint and rewards DIY approaches.  If you’re an IT development shop ELK is an excellent choice.
For large and sophisticated firms, especially financial institutions, energy companies, defense firms and the like, Splunk is cheaper in the long run.


Splunk Support vs. ELK Support

Part of the total value comparison between Splunk and ELK is support, both from the vendor and the surrounding community.
“Wait, ELK support isn’t free?”
If you’re going to implement and run mission-critical IT monitoring tools, then proper support level agreements (SLAs) and enterprise-level engineering processes are mandatory.
Splunk is a fully integrated indexing and analytics package with enterprise-level support from both Splunk, Inc. and the huge Splunk developer community. Buying a Splunk license provides these critical support items. Splunk has supported thousands of installations worldwide.

ELK now offers paid support, SLAs, etc. These services are not free and essentially push ELK into the “freemium” model.  Not a bad move on ELK’s part, just unproven.  ELK paid support doesn’t yet have experience supporting hundreds of large corporations.
“Is our IP protected?”
Splunk has operated a secure support infrastructure for years.
The ELK open source community is very active with support but there are no data confidentiality, security or IP protections when sharing an issue with the ELK community. This lack of IP protection does not pass stringent financial, healthcare or defense industry requirements.
“What if something goes really wrong?”
ELK = elasticsearch + logstash + kibana. These are three different businesses (aka “projects”) which have a symbiotic relationship.  Relying on three different open source businesses for one solution carries significant operational and legal risk.
Splunk, Inc. is a major publicly listed company with deep pockets and well-defined operations. They are a viable and stable corporate partner presenting little risk.
“Can I trust the vendor’s partners to do solid work?”
Splunk has a rich network of preferred domain experts like Risk Focus that must go through intensive training and certification. The installation and app development work Splunk partners do is generally reliable and meets mission-critical engineering standards.
ELK consultancies are emerging, but the community does not yet have a disciplined and well-trained network of implementation partners held firmly to common standards.  ELK has no leverage over independent ELK implementation consultants to follow their standards.  Buyer beware.
We are open source proponents, but these issues merit serious consideration when dealing with complex regulated institutions like banks, brokers, asset managers and exchanges.



Splunk Apps vs. ELK Plugins

One of the major areas of value when adopting an enterprise-wide technology is the ability to easily integrate it with your existing IT infrastructure and applications.
Integration is normally achieved through pre-built adapters, ETL components, apps, plugins and the like.   The more pre-built adapters available, the faster a company can roll out the technology and obtain value without spending significant sums on development, testing and integration.

The Splunk App Store is a rich source of free and paid apps which you can use off the shelf or customize.  Splunk’s app selection covers most generic enterprise software (Outlook, SQL Server, Oracle, etc.), messaging solutions (TIBCO), databases, servers, firewalls, security applications, gateways, etc.
ELK has far fewer pre-built apps (plugins).  Elasticsearch has a number of them.  Most are scattered around different vendors.  There is no comprehensive certification process for ELK plugins.  ELK is in major catch-up mode here and adopting ELK means lots of DIY development to extend it across the enterprise.
To our knowledge there are no financial services app specialists for ELK and very few, if any, in most other industry verticals.  The ELK community is growing fast so the landscape of specialists will change, but for now Splunk has substantially more penetration across industry verticals.

No comments:

Post a Comment

bee-social